École des Hautes Études en Sciences Sociales Ecole doctorale de l’EHESS Nom du centre de recherches Doctorat Discipline : Philosophie FERNANDEZ VELASCO, PABLO A Unified Theory of Disorientation Thèse dirigée par: Roberto Casati Date de soutenance : le 23 avril 2021 Rapporteurs 1 Marcella Schmidt di Friedberg, University of Milano– Bicocca 2 Christopher Peacocke, Columbia University Jury 1 Barbara Tversky, Columbia University 2 Jérome Dokic, EHESS 3 Roberto Casati, EHESS 4 Marcella Schmidt di Friedberg, University of Milano– Bicocca 5 Christopher Peacocke, Columbia University
Transcript
École des Hautes Études en Sciences Sociales
Ecole doctorale de l’EHESS
Nom du centre de recherches
Doctorat
Discipline : Philosophie
FERNANDEZ VELASCO, PABLO
A U n i f i e d T h e o r y o f D i s o r i e n t a t i o n
Thèse dirigée par: Roberto Casati
Date de soutenance : le 23 avril 2021
Rapporteurs 1 Marcella Schmidt di Friedberg, University of Milano– Bicocca 2 Christopher Peacocke, Columbia University
Jury 1 Barbara Tversky, Columbia University 2 Jérome Dokic, EHESS 3 Roberto Casati, EHESS 4 Marcella Schmidt di Friedberg, University of Milano– Bicocca 5 Christopher Peacocke, Columbia University
— 2 —
R E M E R C I E M E N T S
This book has benefitted enormously from the help and advice of a large number of people. Very particular thanks go to Roberto Casati for his encouragement, advice and insight. Also to Slawa Loew, who collaborated with me in producing some of the work that has been adapted into two chapters of the present monograph. To Rory Madden, Hugo Spiers, Andy Clark, Elisabeth Pacherie, José Araya, Valérian Chambon, Andy Clark, Marco Inchingolo, Bastien Perroy, Valeria Giardino, Solène Le Bars, Elisabeth Pacherie, George Neish, Takuya Niikawa and Nura Sidarus, Olivier Massin, Giuseppe Attoma, Jérôme Dokic, Joëlle Proust, Virginie Schmidt and Agostino Pinna Pintor for their comments on previous drafts of this work and for general discussion about the topic; to the team at the Disorientation Gr; to the presenters and participants of the Disorientation Seminar at Institut Jean Nicod; to the team at the UCL Spatial Cognition Lab; to the European Cultural Foundation and the Royal Geographical Society that funded the fieldwork behind some of this work; to Edwin Hutchins and the participants of the Cognition in the Wild workshop that took place at the EHESS in 2018; to Jaanika Vider for all of her work in our shared fieldwork in Siberia; to the whole community in Evenkia for their hospitality, their kindness and their insight; to the artists Emilie Berry, Maxime Scocard, Annabelle Vailant, Filippo Fabbri, Saida Makhmudzade, Samuel Garcia-Filastre, Andrea Fortier and Uliana Leonova, who have helped me explore some of these philosophical questions from a radically different angle; to Fiona Zisch and to the graduate students at the Interactive Architecture Lab of the UCL Bartlett School of Architecture. I am forever grateful to Anya Gleizer, who has been my relentless companion in the many adventures that fueled my interest in disorientation in the first place, and to my family for their love and support. Some of the chapters in this monograph reproduce or draw on material from the following articles:
Fernandez Velasco, P., Loev, S. Affective experience in the predictive mind: a review and new integrative account. Synthese (2020). Fernandez Velasco, P & Casati, R. (2020). Subjective Disorientation as a Metacognitive Feeling. Spatial Cognition and Computation. Fernandez Velasco, P & Casati, R. (2020). The many faces of disorientation: A response to Daniel R. Montello. Spatial Cognition and Computation. Fernandez Velasco, P. and Casati, R. (2021). Making and breaking our shared world: A phenomenological analysis of disorientation as a way of understanding collective emotions in distributed cognition. In The Politics of Emotional Shockwaves. Palgrave Macmillan. Fernandez Velasco, P. (2020). Disorientation and Self-consciousness: A Phenomenological Inquiry. Phenomenology and the Cognitive Sciences. Fernandez Velasco, P. & Casati, R. (2019). Disorientation and GIS-Informed Wilderness Search and Rescue. In The Philosophy of GIS (pp. 241-251). Springer, Cham. Fernandez Velasco, P & Loev, S. (in preparation). An Active Inference account of affect and epistemic activity.
— 3 —
R É S U M É E T M O T S C L É S
Résumé Il existe un grand nombre d'études sur la désorientation spatiale, allant de la
géographie aux neurosciences, mais il n'existe pas de compréhension unifiée du phénomène. En
réponse à cette hétérogénéité, la présente thèse de doctorat propose une théorie unifiée de la
désorientation spatiale.
La désorientation est, au fond, un phénomène subjectif, et c'est en accordant une attention
appropriée à ses aspects subjectifs que nous pourrons atteindre la charactérisation recherché. Une
recherche empirique qualitative - telle qu'un corpus de rapports subjectifs recueillis par Qualtrics et
un travail de terrain ethnographique avec des chasseurs indigènes Evenki en Sibérie arctique - guide
à la fois l'analyse conceptuelle et l'enquête phénoménologique de cette thèse. L'analyse
phénoménologique de la désorientation révèle le lien entre notre sens de l'espace et notre sens de
possibilité. Au cours de notre orientation quotidienne, il y a une intégration continue des
représentations spatiales indexées et non-indexées de notre environnement. Il en résulte que dans
notre expérience standard d'un environnement connu, il y a toujours une configuration spatiale hors
de vue qui se profile au-delà de l'horizon de notre expérience et qui encadre cet horizon. Lorsque
nous sommes désorientés, nous perdons ce sens de la localisation de notre environnement hors de
vue par rapport à notre position actuelle. Le plus souvent, lors d'une désorientation, nous nous
trouvons dans un espace à la fois appauvri et oppressant. Nous perdons non seulement le sens de la
structure de notre environnement, mais aussi celui de nos possibilités dans cet environnement. Il en
résulte généralement de la confusion, de l'anxiété et de l'impuissance.
L'analyse conceptuelle et phénoménologique de cette monographie contribue à faire avancer
sa thèse centrale, à savoir que la meilleure façon de caractériser la désorientation est de la
considérer comme un sentiment métacognitif. Ici, la façon de comprendre les sentiments
métacognitifs (comme le sentiment de savoir ou le sentiment de familiarité) est de les considérer
comme des expériences affectives concernant les propres états, processus ou capacités mentaux du
sujet. Dans notre cas, un sous-système cognitif évalue et régule les processus de navigation actifs du
sujet, ce qui se traduit par l'expérience de désorientation.
— 4 —
La désorientation ne fonctionne pas seulement comme un moyen d'évaluer les représentations
spatiales dans la tête d'un individu. La désorientation joue un rôle dans l'évaluation des processus de
navigation qui peuvent être distribués au-delà de l'individu, sur des artefacts cognitifs (par exemple,
une carte ou un GPS), l'environnement (par exemple, un système de signalisation des sentiers) et
même des groupes de personnes (par exemple, un guide ou un compagnon de chasse). Lorsqu'il
existe des raisons de ne plus faire confiance au système de navigation en place, la désorientation
apparaît pour réguler la situation de manière dynamique. Étant donné que la désorientation se situe
à l'intersection de tant d'aspects cruciaux de la cognition humaine, lorsque nous jetons une lumière
sur le phénomène, nous obtenons également un aperçu de certains de ces aspects. Nous voyons
également comment les sentiments (la désorientation en est un excellent exemple) nous guident afin
que nous puissions nous intégrer dans les vastes écosystèmes cognitifs que nous habitons. Nous
voyons également comment les processus cognitifs distribués modulent notre expérience de
l'espace, et comment cette expérience de l'espace structure à son tour notre expérience du monde et
de nous-mêmes. Et bien sûr, nous voyons ce qui se passe lorsque cette structure familière
s'effondre.
Mots clés Désorientation, Phénomenologie, Metacognition
— 5 —
A B S T R A C T A N D K E Y W O R D S
Abstract There is a large body of literature investigating spatial disorientation —ranging from
geography to neuroscience—, but there is no unified understanding of the phenomenon. In response
to this heterogeneity, the present doctoral thesis advances a unified theory of spatial disorientation.
Disorientation is, at heart, a subjective phenomenon, and it is by devoting proper attention to
its subjective aspects that we can reach the sought-for integrative account. Qualitative empirical
research —such as a corpus of subjective reports collected through Qualtrics and ethnographic
fieldwork with Evenki indigenous hunters in arctic Siberia— guides both the conceptual analysis
and the phenomenological investigation in this thesis. The phenomenological analysis of
disorientation reveals the link between our sense of space and our sense of possibility. During our
everyday orientation, there is an ongoing integration of indexical and non-indexical spatial
representations of our environment. The result is that in our standard experience of a well-known
environment there is always an out-of-sight spatial configuration looming beyond the horizon of our
experience and framing that horizon. When we are disoriented, we lose this sense of where our out-
of-sight surroundings are located with respect to our present position. Most often, during
disorientation we find ourselves in a space that is both impoverished and oppressive. We lose not
only our sense of the structure of our environment, but also of our possibilities within that
environment. This commonly results in confusion, anxiety and helplessness.
The conceptual and phenomenological analysis in this monograph help advance its central
thesis, which is that the best way to characterize disorientation is as a metacognitive feeling. Here,
the way to understand metacognitive feelings (like the feeling of knowing or the feeling of
familiarity) is as affective experiences concerning the subject’s own mental states, processes or
capacities. In our case, a cognitive subsystem evaluates and regulates the subject’s active
navigational processes, and this results in the experience of disorientation.
— 6 —
Disorientation works not only as a way of evaluating the spatial representations inside an
individual’s head. Disorientation plays a role in evaluating navigational processes that can be
distributed beyond the individual, onto cognitive artefacts (e.g. a map or a GPS), the environment
(e.g. a trail signage system) and even groups of people (e.g. a guide or a hunting companion). When
there are reasons to withdraw confidence in the ongoing navigational system, disorientation
emerges to regulate the situation dynamically. Because disorientation is at the intersection of so
many crucial aspects of human cognition, when we throw a light onto the phenomenon, we gain
insight into some of those aspects as well. We see also how feelings (disorientation being a prime
example) guide us so that we can fit within the broad cognitive ecosystems that we humans inhabit.
We see as well how distributed cognitive processes modulate our experience of space, and how that
experience of space in turn structures our experience of the world and of ourselves. And of course,
we see what happens when that familiar structure falls away.
Figure 2. Illustration of Scenario 2 ……........................................................................................ 57
— 9 —
I N T R O D U C T I O N
Studying disorientation is studying how, through our bodies, culture and technology, we humans are connected to our environment and what happens when this connection is weakened or severed. What happens, of course, depends again on our environment, bodies, culture and technology: it is not the same to become lost in the woods as in the mall, not the same to be unable to interpret a map as to experience a GPS glitch. On some occasions, getting lost strikes terror into our hearts, and we panic and start to wander aimlessly. In others, getting lost is a red light that tells us to stop on our tracks, as we take time to reconsider our bearings and devise orienteering strategies. The world around us becomes at times uncanny, unfamiliar or dangerous when we get disoriented and, at times, disorientation is exciting and refreshing — an invitation to explore, to leave behind nagging desires for control and certainty, and to embrace instead a more spontaneous relationship with our surroundings. Getting lost shapes our consciousness, not only by transforming our perception of the world around us, but by transforming our sense of who we are in that world and what possibilities are open to us within it. It might seem at first that in our highly technological world of GPS units and navigation apps, disorientation is a rare occurrence that should be looked at almost with nostalgia. Far from that, our blind reliance on technology can itself be a cause for disorientation. And our modern globalised world of hyperinformation and rapid changes can be most disorienting. Shopping malls and large underground transport hubs, despite the abundance of maps and signage, are some of the places where it is easiest to get lost, precisely because of their placelessness; a mall looks just like any other mall and a coffee chain outlet just like any other outlet of that same coffee chain. In these post-modern environments there is virtually no sense of place that can be developed and no distinguishable landmarks on which to ground our orientation. Nowadays, disorientation is, in fact, pervasive, and what makes studying it challenging is not its rarity but its ubiquity and its multifaceted nature. One can study disorientation through the lens of neuroscience or through the lens of human geography, and with a focus ranging from search and rescue operations to the architecture of public spaces. As of today, the myriad fields that deal with disorientation have not been properly connected and what we have is not a unified picture of the phenomenon, but a set of fragmented takes on it. What is needed, or so I shall argue in this monograph, is a unifying characterisation at a fundamental level, from which a theory of disorientation can emerge to the benefit of the different disciplines concerned. Disorientation is, at heart, a subjective phenomenon, and it is by devoting proper attention to its subjective aspects that we can reach the sought-for characterisation. Accordingly, a large part of the present work is a phenomenological investigation. Throughout the monograph, this central phenomenological approach is put into communication with other disciplines, such as neuroscience, geography or ethnography. This interdisciplinary approach is in line with the multifariousness of the current state of disorientation research, and it both constraints and guides the conceptual work at
— 10 —
different explanatory levels (e.g. phenomenal, neural, computational, behavioural). In return, the result is a theory that can be adapted to provide explanations and predictions for a broad array of disciplines. A phenomenological analysis of both the negative and the positive aspects of disorientation reveals the link between our sense of space and our sense of possibility. During our everyday orientation, there is an ongoing integration of non-indexical spatial representations of our environment (i.e. coded in allocentric frames of reference) and of indexical representations of that environment (i.e. coded in egocentric frames of reference). The result is that in our standard experience of a well-known environment there is always an out-of-sight spatial configuration looming beyond the horizon of our experience and framing that horizon. If we know Paris well, when we are walking around Jardin de Luxembourg we will have a sense of where the exits of the park are, and even of where Notre Dame (and Port Royal, Montparnasse, the Pantheon…) is, so that we could point at it without explicitly imagining a map. When we are disoriented, we lose this sense of where our out-of-sight surroundings are located with respect to our present position. Most often, during disorientation we find ourselves in a space that is both impoverished and oppressive. We lose not only our sense of the structure of our environment, but also of our possibilities within that environment. This commonly results in confusion, anxiety and helplessness. However, there are some rare occasions in which disorientation can be a cause for joy and excitement. In these occasions, the falling away of the standard structure with which we apprehend space becomes an invitation for relating to our surroundings differently. We develop a feeling of liberation as we let ourselves be guided by the attractions of the terrain. This two-headed nature of getting lost is both a source of wonder and a central clue in our understanding of the phenomenon. The conceptual and phenomenological analyses in this monograph help advance its central thesis, which is that the best way to characterize disorientation is as a metacognitive feeling. Here, the way to understand metacognitive feelings (like the feeling of knowing, the tip-of-the-tongue state or the feeling of familiarity) is as affective experiences concerning the subject’s own mental states, processes or capacities. In our case, a cognitive subsystem evaluates and regulates the subject’s active navigational processes, and, in certain conditions, this results in the experience of disorientation. The proposed theory is nested within two existing frameworks: Predictive Processing and Distributed Cognition. Predictive Processing considers prediction to be the fundamental activity of the human mind. According to Predictive Processing, the brain makes hypotheses about the world based on previous experience and these hypotheses get updated with error coming from sensory sheets in a feedback loop that operates at different timescales and levels of abstraction. The central idea behind Distributed Cognition is that cognitive processes are distributed not only within our skull, but also across a cognitive ecosystem that emerges from the interaction of many elements such as our body, a group of agents or cognitive artefacts. A combination of the two gives us a view in which agents operate within cognitive ecosystems through distributed cognitive processes that optimise prediction success over time. Metacognitive feelings such as disorientation, I defend, are a central element in the regulation of the different elements that are involved in distributed cognitive processes, through evaluative control in a fine-tuned prediction-updating loop. There is no need to give more details about the theory at this stage. Suffice to say that, with this theory, we see how disorientation works not only as a way of evaluating the spatial representations inside an individual’s head. Disorientation plays a role in evaluating navigational processes that can be distributed beyond the individual, onto cognitive artefacts (e.g. a map or a GPS), the environment (e.g. a trail signage system) and even groups of people (e.g. a guide or a hunting companion). When there are reasons to withdraw confidence in the ongoing navigational system, disorientation emerges to regulate the situation dynamically. Because disorientation is at the
— 11 —
intersection of so many crucial aspects of human cognition, when we throw a light onto the phenomenon, we gain insight into some of those aspects as well. We see also how feelings (disorientation being a prime example) guide us so that we can fit within the broad cognitive ecosystems that we humans inhabit. We see as well how distributed cognitive processes modulate our experience of space, and how that experience of space in turn structures our experience of the world and of ourselves. And of course, we see what happens when that familiar structure falls away.
I — METHODOLOGY The present work engages with many contemporary debates in analytic philosophy (e.g. the links between self-location and self-consciousness), anthropology (e.g. mind map theory vs practical mastery theory of wayfinding) and cognitive science (e.g. how cognition is distributed beyond the individual), to name but a few disciplines. The notions, levels of explanation and even what is considered a valid explanation differ among disciplines. As a result, walking between disciplines is always walking a fine line. Here, phenomenology can serve as a way to find common ground. A good understanding of the phenomenon can give us insight and guide interdisciplinary inquiry. This, together with rigorous conceptual analysis can help advance our knowledge of a subject matter that is destined to require interdisciplinary efforts. The phenomenological and conceptual analysis in this monograph builds on a series of reports about individuals getting lost in a variety of situations. Some of these reports are historical. In chapter 2, I draw on the 19th century corpus of reports that Alfred Binet collected to understand the phenomenon of getting ‘turned around’, a subclass of disorientation experiences with a very interesting phenomenology. In chapter 3, I draw on the report of the early 20th century anthropologist Maria Czaplicka, who writes about how her Evenki native guide got them lost in the Siberian taiga in the middle of a snowstorm. Other reports are derived from contemporary sources. Kenneth Hill collected a series of reports from lost hunters in order to categorise the behaviour of disoriented subjects, which informs some of the conceptual work in this monograph. Another useful set of reports comes from astronauts that get ‘turned around’ (along the vertical instead of the horizontal axis!) in microgravity environments, leading to a most striking phenomenology. The vast majority of contemporary reports come from a survey that the Disorientation team at Institut Jean Nicod (of which I am a member of) started back in 2017 with the goal of forming a corpus of diverse experiences of disorientation. In said reports, subjects expound on their experiences of disorientation, conveying details about the situation that prompted the episode and the emotions associated with it. A final set of reports comes from ethnographic work among semi-nomadic Evenki natives in arctic and subarctic Siberia. In 2019, I embarked on two expeditions (one in the winter and one in the summer) to the Evenkia region of Krasnoyarski Krai in Russia. Evenki hunters and reindeer herders are famous for their wayfinding skill, and a large part of my fieldwork consisted in learning about their navigational methods and how these practices were embedded within a larger cognitive ecosystem. Another aspect of Evenki culture that I wanted to learn about was the phenomenology of space and self that they experience during wayfinding. In particular, I was interested in the feeling that they call ‘manakan’, a sense of autonomy and of engagement with their environment that echoes the phenomenology of the positive manifestations of disorientation. The reports that I collected during these two fieldwork seasons in Siberia inform the comparative study in chapter 3. In the earlier chapters of the dissertation, subjective reports are often used to clarify the phenomenology of disorientation and guide a phenomenological analysis. In the chapter in which the central claim is advanced, subjective reports are also used as scenarios that delimit what should
— 12 —
and should not be considered disorientation. Using first-person data to establish a theory that can be applied to explain third-person data poses a challenge, because there is an explanatory gap when it comes to relating first-person to third-person data. There are at least two possible ways to face this challenge. The first one is to use the first-person data coming from the phenomenology of disorientation to bring forth testable hypotheses about cognition. The second one is to use the first-person data coming from the phenomenology of disorientation to uncover links between disperse existing empirical data, and then use this empirical data to support hypotheses about cognition. The approach that I follow in this work is a combination of the two. In other words, the phenomenological analysis of disorientation gives us insight into the phenomenon, which helps us re-interpret existing empirical work. And whatever conclusions are reached through this process will generate claims about cognition (e.g. that disorientation is a metacognitive feeling) that are empirically testable using third-person data.
II — OUTLINE
To arrive at the desired unifying theory of disorientation, I proceed in the following way: Chapter 1 introduces the different disciplines that deal with disorientation and argues for the importance of developing a unified theory of the phenomenon. I then use Search and Rescue as a case study to show how this theory can advance research in different fields. Chapter 2 explores the phenomenology of disorientation and its relationship with self-consciousness. It starts by discussing previous literature on the links between self-location and self-consciousness and proposes a distinction between minimal self-location (which requires only an egocentric frame of reference) and integrated self-location (which requires the integration of egocentric and allocentric frames of reference). The aim of this chapter is to use this distinction between minimal and integrated self-location to deepen our understanding of spatial disorientation. A phenomenological analysis of different disorientation episodes reveals that during paradigmatic disorientation cases, the body-space shrinks, and the horizon of experience becomes more uncertain, leading to anxiety and to a feeling of unfamiliarity. The main claim of this chapter is that during disorientation, a destabilization of integrated self-location results in a diminished form of self-consciousness. Chapter 3 explores the positive aspects that are sometimes associated with experiences of disorientation. To try to make sense of this apparent paradox, I turn to another apparent paradox: the case of Evenki reindeer herders in northern Siberia, who are reported to never get lost yet experience the landscape in a way that shares some of the positive aspects of disorientation episodes. Here, I draw on secondary literature and on my own interactions with Evenki reindeer herders to establish the following hypothesis: a deep knowledge of their landscape and a remarkable wayfinding capacity means that Evenki navigation is not largely driven by allocentric representations. This means that whether or not allocentric and egocentric frames of reference are integrated is not a central factor in the navigational processes involved in typical Evenki spatial behaviour, and therefore the destabilisation of integrated self-location is not something that surfaces in the conscious experience of Evenki reindeer herders. The Evenki never get lost not because they always know precisely where they are in respect to a large allocentric frame of reference, but because their interactions with their environment are structured in such a way that they can walk relying only on egocentric frames of reference and still be able to always find their way. This means, in turn, that their conscious experience of their environment (and of themselves within that environment) is largely driven by egocentric and not by allocentric frames of reference, which
— 13 —
results in an experience of the large-scale environment as underdetermined. This, however, does not lead to anxiety, but to a sense of novelty and possibility. Taking this analysis to reflect on the positive aspects of disorientation experiences, I conclude that the uncertainty of the surrounding space during disorientation can lead not only to anxiety, but, alternatively, to an experience of the environment as novel and rife with possibility. Chapter 4 builds on the corpus of reports that the Disorientation team have collected to advance a unified characterization of disorientation. This chapter first distinguishes between the objective condition of being lost and the subjective condition of disorientation. The central claim of Chapter 4 (and of the overall monograph) is that disorientation is a metacognitive feeling that evaluates and regulates active navigational processes. This characterization is compatible with both third-person empirical data and first-person data on the subjective experience of disorientation, and it serves as a central point to advance a theory of the phenomenon. Chapters 5 and 6 aim to offer an account of affective experiences within Predictive Processing. Chapter 5 begins by outlining a set of common features of affective experiences (or feelings) that a Predictive Processing theory should aim to explain: feelings are conscious, they have valence, they vary in their degree of arousal, they motivate behaviour, and they are intentional states with particular and formal objects. I then review existing theories of affective experiences within Predictive Processing and delineate two families of theories: Interoceptive Inference Theories (which state that feelings are determined by interoceptive predictions) and Error Dynamics Theories (which state that feelings are determined by properties of error dynamics). The chapter then highlights the strengths and shortcomings of each family of theories. Chapter 6 builds on the review of the previous chapter to provide a synthesis of the existing families of theories of affective experience within Predictive Processing: the Affective Inference Theory. The Affective Inference Theory claims that valence corresponds to the expected rate of prediction error reduction. In turn, the particular object of a feeling is the object predicted to be the most likely cause of expected changes in prediction error rate, and the formal object of a feeling is a predictive model of the expected changes in prediction error rate caused by a given particular object. Finally, the Affective Inference Theory shows how affective experiences bias action selection, directing the organism towards homeostasis and towards optimal levels of uncertainty in order to minimise prediction error over time. In the last section of the chapter, I employ the Affective Inference Theory to account for how metacognitive feelings emerge and how they guide behaviour. Chapter 7 takes disorientation as a case study to elucidate the role that affective states play in distributed cognitive processes. The chapter fist introduces Distributed Cognition and recent situated approaches to emotion. Then, I introduce some cases of distributed cognition in navigation and look at the case of spatial disorientation to explore the role of affect in navigation within the broader context of situated cognition. The central claim in this chapter is that in distributed cognitive processes, affective states serve as a form of evaluative regulation that contributes to the modulation of a distributed cognitive process and to the eventual switch between distributed cognitive processes. Although they are connected thematically and they reference one another, chapters 1 to 7 do not depend too heavily on each other, in the sense that later chapters do not rely strongly on arguments or claims made in previous chapters. The exception to this is chapters 5 and 6, which work as a tandem. This level of independence is particularly important because chapter 5-6 and chapter 7 do not argue directly for Predictive Processing or for Distributed Cognition, respectively. They explain what the theoretical frameworks consist of and they develop an account of affective states within those frameworks. I point the reader to sources that defend the frameworks directly, but if a reader
— 14 —
has independent reasons against one of the frameworks, they will find that they can read the rest of the monograph without encountering strong commitments to that framework. Chapter 8 weaves together the work developed in the rest of the monograph. The chapter begins by synthesising the accounts of affective states developed in chapters 5, 6 and 7 in order to explain metacognitive feelings. Then, this work is used to account for the functioning of disorientation, adding flesh to the theoretical bone introduced in chapter 4. This new theory of disorientation is used to explain the phenomenology of disorientation, both in its negative and positive manifestations, as it was presented in chapters 2 and 3. Finally, I show how the theory can help advance the different disciplines concerned with the phenomenon that had been introduced in chapter 1. My hope is that after reading chapter 8, the reader will be convinced that this monograph has accomplished what it set itself to do at the outset — to provide a theory of how we humans are connected to our environment and of what happens when this connection falters.
— 15 —
1 . T H E N E E D F O R A U N I F I E D T H E O R Y O F
D I S O R I E N T A T I O N
I — REVIEW OF THE DISORIENTATION LITERATURE
Again, something dark appeared in front of him. Again, he rejoiced, convinced that now it was certainly a village. But once more it was the same boundary line overgrown with wormwood, once more the same wormwood desperately tossed by the wind and carrying unreasoning terror to his heart.
-Tolstoy. Master and Man Tolstoy’s story of a disoriented man in a snowstorm is not far off from actual reports of lost person behaviour. After realising that he’s been walking in circles, Vasili Andreevich tries to keep calm, but assailed by a mounting panic, he is unable to collect himself and stay put. Instead, he starts to wander aimlessly, which is a typical lost person behaviour (Hill 1998). Kenneth Hill conducted over a hundred interviews with subjects who had become lost in the wilderness and constructed a list of typical disoriented behaviour that has now become a central reference in the study of lost person psychology. Hill claimed that when lost, people followed (at least) one of the following patterns and strategies: random traveling (often caused by confusion and high emotional arousal), route traveling (i.e. following a given route in the hope of finding something familiar), direction traveling (trying to follow a given direction), route sampling (using an intersection as a base for exploring different routes), direction sampling (using a visible landmark as a base for exploring different directions), view enhancing (aiming for a high position in order to gain visibility), backtracking, using folk wisdom (e.g. follow streams to civilisation) and staying put (which according to Hill is the most effective strategy if a search and rescue operation is triggered). There has been some work in search and rescue theory relating this taxonomy of lost person behaviour with different categories of lost subjects (e.g. hunters are more likely than others to use direction sampling, while hikers are more likely to use route following). Nevertheless, there isn't any work relating the taxonomy to a broader conceptualisation of disorientation (e.g. how disorientation arises and what makes subjects follow one method over another) or to the work in other fields such as neuroscience. This lack of connection is to be expected when it comes to disorientation, because the literature on disorientation is not a unified body of research. Rather, there are many fields that deal with the phenomenon (Schmidt di Friedberg 2017). The historical record is rich and variate, because coping with difficulties in orientation has been a major challenge in the history of navigation (Huth 2013). Most of the recent data on human disorientation in real-life settings has been collected for the science of search and rescue operations through the study of lost person behaviour (Koester 2008; Sava et al 2016; Lin and Goodrich 2010). Another large body of
— 16 —
work comes from clinical psychology and from neurology, through the study of conditions such as topological disorientation or Alzheimer’s disease (Henderson, Mack and Williams 1989; Monacelli et al. 2003). The patients who suffer from topological disorientation have selectively lost their ability to find their way within large environments (Aguirre and D’Esposito 1999). Work on disorientation has also been done in cognitive science and in psychology, including animal and developmental studies (see Waller and Hodgson 2006 for an example; see Dudchenko 2010 for an overview of the literature). However, there is still a missing theoretical link connecting the work in these different fields. In an effort to establish a link between the behavioural psychology and the neuroscience of disorientation, Dudchenko holds that humans, unlike other animals, need vision to stay oriented (Dudchenko 2010). This makes sense because humans use landmarks to update one’s orientation and position within a cognitive map (Scholl 1987; Knierim and Hamilton 2011). The notion of cognitive maps was originally developed by Tolman as a hypothesis to explain rat behaviour in labyrinths without reducing the said behaviour to stimulus-response connections. Tolman supported the view that rats construct something similar to a field map with results from different experiments, including shows of initial latent learning that gets activated when reward is offered later (Blodgett 1929), rats sampling of the environments (Tolman 1938), and rats finding shortcuts that they had not learned through stimulus-response conditioning (Tolman 1948). The discovery of place cells in the rodent hippocampus further supported the idea of a cognitive map (O’Keefe and Dostrovsky 1971), which has now become central in the study of spatial cognition. Place cells fire as a function of the rodent’s spatial location. Other cells that may play a role in navigation are grid cells (which fire in a hexagonal grid that corresponds with the environment floor, where the grid module is correlated with the size of the animal; Hafting et al. 2005), head direction cells (which fire depending on head orientation relative to a landmark; Ranck 1985; Taube et al. 1990) and boundary vector cells (which fire when the rodent, moving in a certain direction, gets to a specific distance from a perceived boundary element of the environment; Barry et al. 2006). An observation is in order. Different classes of spatial features are coded by different cell systems, as per the following table: Cell type Geometric feature coded Example
Place cells Individual locations The entrance of the maze, the fork, etc.
Grid cells Metric properties Distance covered based on number and size of grid modules
Head direction cells Angles Azimuth of a landmark relative to body’s central plane of symmetry
Boundary vector cells Topological properties The middle of the room, close to the wall
This suggests that the spatial representation system could work in a constraint satisfaction mode, looking for solutions that take into account whichever topological, metric, angular and location representations are available at a given moment. Of course, an open question is if the same type of cells and the same type of cognitive map exists in humans. Although there are some limitations, there is mounting empirical evidence in support of
— 17 —
the idea that work on rodent spatial cognition can be extrapolated to humans to a substantial degree (see Epstein et al. 2017 for a review). For instance, a much-quoted fMRI study of Taxi drivers in London show that their right posterior hippocampi become larger as a result of the years-long training they undertake to learn the map of London streets (Maguire et al. 2000). If the underpinnings of human spatial cognition do not differ substantially from those of rodent spatial cognition, then the hippocampus can store different maps for different environments, and even represent different conditions of the same environment. Different place cells would then fire for different environments, following a ‘global remapping’ as the subject changes environments. When the same environment changes conditions, the place cells that fire stay the same, but the rate at which they fire changes (Colgin, Moser and Moser 2008). Finally, when it comes to navigation, the grid cell network computes the direction and distance to the navigational goal (Kubie and Fenton 2012), while the hippocampus and enthorinal cortex work together to calculate the optimal path to the goal (Bryne, Becker and Burgess 2007). It is beyond the scope of this monograph to explore in depth the neuroscientific literature on spatial cognition, but the above shows that there have been important advances that can help us understand how humans orient themselves. The neuroscience of navigation can help us understand how the brain makes orientation possible, and it can give support to concepts that can be useful in conceptualising navigation, such as the notion of a cognitive map. However, learning about the neuroscientific underpinnings of navigation can only take us that far if what we want to learn about is disorientation. What we need to understand is not only how navigational processes are implemented, but the limitations of those processes, and most importantly, how those processes are evaluated and regulated — we need a comprehensive understanding of how disorientation emerges and how it guides behaviour. As such, studying disorientation requires studying not only human internal computational processes but also human experience, human behaviour and human cognitive ecosystems. In other words, it requires an interdisciplinary approach. And at present, there are virtually no links between research in neuroscience and search and rescue theory, or between architecture and the study of topological disorientation, to give but a few examples. The result of the current disconnect in the study of disorientation is that there is no unified theory that can draw on the multifarious research that deals with the phenomenon. In what follows, I will discuss the conceptual issues that stand in the way of a unified characterisation, and I will use the field of Wilderness Search and Rescue (WiSAR) as a case study of how a unified theory can advance our understanding of disorientation and thus be beneficial for a broad variety of disciplines.
II — CONCEPTUAL ISSUES WITH THE STUDY OF DISORIENTATION
There are several impediments that keep the different fields concerned with disorientation from associating. An important one is that these fields operate at different levels of analysis. Even within neuroscience, where most explanations appertain to the neural level, there is a chasm between the single-cell study that is possible in non-human cognition and the non-invasive neuroimaging studies done in human cognition. Moreover, functional magnetic resonance imaging cannot be done while subjects travel physically through space, which restricts investigation to non-ecological, purely represented settings. Although some lost person psychology relies on first person reports (Hill 1982), for the most part it is concerned with the third-person analysis of behaviour. In the case of Wilderness Search and Rescue (WiSAR) research, the separation from spatial cognition goes one step further, as most of WiSAR work operates at the level of statistical patterns emerging from lost person behaviour (Koester 2008).
— 18 —
Another important element that separates these fields is that they tend to draw from different experimental and study settings. A big part of the work in psychology takes place in non-ecological conditions (e.g. Li, Mou and McNamara 2012). WiSAR theory is based on ecological non-experimental conditions (i.e. analysis of existing cases of lost people), and the work in computational cognitive science is usually simulation-based (e.g. Ruddle, Volkova, Bülthoff 2011). Finally, as mentioned above, neuroimaging studies are stationary, and thus based on virtual reality, spatial recollections or imagined navigation. Consequently, in order to extrapolate results from one field to another, one has not only to extrapolate from one level of analysis to another, but also from one setting to another (e.g. from simulations to the actual behaviour of people lost in the wilderness). A symptom of this divergence among fields is a divergence among characterisations of the phenomenon of disorientation itself. Disorientation could be understood, among others: as being unable to find one’s way (Dudchenko 2010); as the failure of the way-finding process (Golledge 1999); or as not knowing the directions and distances to get to a given point (Rieser 1999). These all seem to point to the same phenomenon, but it should be noted that they each characterise disorientation in a proprietary way: as an issue of ability, failure or ignorance, respectively. Although the above characterisations each capture some of the features of disorientation, they all have certain shortcomings. A central issue here is that while being lost in an objective condition, disorientation is a strongly subjective one. Being lost can be easily characterised as being unable to find one’s way, but this characterisation does not necessarily extend to disorientation. Take the following case, for instance:
Maggie is following a path through the woods that leads to the Dorlcote Mill. The path is very hilly and sinuous, changing directions often, which makes Maggie feel disoriented. Disregarding her disorientation, she keeps going and eventually gets to the Dorlcote Mill.
In this case, Maggie is not lost (i.e. she is able to find her way), but she is disoriented; a distinction that Duchenko’s characterisation fails to capture. And while Dudchenko’s characterisation fails to capture some cases of disorientation, Golledge’s characterisation overshoots, mistakingly identifying certain cases of being lost as disorientation. Let us look at the following example:
Pierre is confident that he is at Place Denfert Rochereau, when in fact he just emerged at the Port Royal crossing, because he skipped one metro stop.
Here, Pierre’s way-finding process failed (he failed to keep track of the stations when he was in the metro), but he doesn't feel disoriented. Golledge would characterise Pierre as being disoriented, but a more precise characterisation would identify Pierre as someone who is lost, not disoriented. Finally, below is a last case that raises some problems for Rieser’s characterisation:
John is a tourist in Paris. He is outside the Metro stop Odéon and he wants to go to walking to Place de la Concorde. He is told that if he follows Boulevard St. Germain and then crosses the Seine he will get there. Boulevard St. Germain curves north as it advances west, but this is not a problem for John. He just follows the street without knowing the exact direction of Concorde. He ends up at the Seine, crosses the river and is at Place de la Concorde.
In the case above John doesn't know the direction or the distance of his destination, but he is neither lost nor disoriented. Overall, there are many elements related to disorientation, such as knowledge of distances and directions, the way-finding process or navigational ability. However, casting disorientation in such terms overlooks its subjective aspects. The difficulty is that we need to
— 19 —
characterise disorientation in non-subjective terms in a way that captures its subjective aspects. This is an issue that will be revisited in chapter 4. Bridging the gap among these different fields is not only an empirical endeavour, but a conceptual one. An important first step is a characterisation of disorientation that can be effectively applied to different fields. As for the gap between levels of analysis, it can be bridged indirectly. That is, results at one level can be used to speculate and produce hypotheses at other levels. Ultimately, one would want to have an overall theory of disorientation that draws on the progress from the variety of fields dealing with disorientation into a unified conceptualisation of the phenomenon that can serve to generate testable hypothesis for different settings and levels of analysis. At first, it might seem that such a conceptual endeavour is overtly theoretical and therefore somewhat divorced from practical applications or from empirically oriented research. I would argue that the opposite is true. Wherever there is an apparently non-theorical pursuit in the study of disorientation there tends to be an implicit theory of disorientation underlying its functioning and its rationale, so that improving said implicit theory will improve the workings of the empirical or practical activity as well. An excellent case in point is search and rescue operations. WiSAR science is probably the most practical application of research into lost person behaviour, so in the remainder of this chapter, I will look at recent advances in WiSAR modeling and show how they depend on an underlying disorientation theory and how novel conceptualisations of disorientation can contribute to WiSAR models that generate better predictions.
III — WISAR MODELS When a person becomes lost in the wilderness, the WiSAR manager faces a difficult challenge, which is to figure out where the different SAR teams are more likely to locate the lost person. The areas that need to be covered are often vast areas of difficult terrain, and the SAR team needs to coincide with the lost person not only in space, but also in time (i.e. if the SAR team and the lost person pass the same place at different times the lost person will not be found). These difficulties should not be overlooked. WiSAR operations often have limited resources, and past two days, or about fifty hours of search, the lost person’s chances of survival decrease significantly (Adams et al. 2007). To guide the search effort, the most common strategy is to develop a probability map stemming from the Initial Planning Point (IPP), which is usually the point where the lost person was last seen or known to be (i.e. relying on cues). Geographic Information Systems (GIS) provides WiSAR the planning tools (such as the ArcGIS planning tool, see Ferguson 2008) to create and update probability maps (Doherty et al. 2014). Here, I will discuss two methods of establishing probability maps: a ring model and a Bayesian model. The ring model, based on mathematical search theory (Koopman 1980), is the most common way of establishing probability areas. The main dependent variable of the ring model is Euclidean distance from IPP. The statistics of past WiSAR cases are analysed to produce four concentric rings stemming from IPP. These rings represent probability areas and are established at the distances equivalent to the probability quartiles 25%, 50%, 75% and then at 95%. That is to say that, within the first concentric ring, the probability of finding the lost person is 25%, within the second concentric ring, the probability of finding the lost person is 50%, and so on. An important finding for WiSAR is that different subject categories correspond to different probability distributions. For example, the 25% probability area of a hiker is much larger than that
— 20 —
of a child, because the hiker can cover bigger distances and tends to follow either trails or directions rather than staying put or walking randomly. Syrotuck first analysed 242 cases from New York and Washington states and established probability distributions for eight subject categories (Syrotuck 2000). Koester has recently created ISRID, a large database unifying thousands of cases from around the world and dividing them into 41 categories based on scenario, age, medical or mental status, and activity (Koester 2008). Although other parameters (such as the type of terrain) can be added, the subject category is the main determinant of the distribution of the probability map that will guide the search efforts. A recent development in WiSAR is Lin and Goodrich’s Bayesian approach (Lin and Goodrich 2010). The main parameters of their model are the topography, vegetation coverage and local slope of the terrain. One of the innovative elements of their model is that it uses expert opinions to determine the probability of the lost subject transitioning from one terrain type to another (e.g. a WiSAR manager’s estimation of the likelihood of a given lost subject moving from a plain to a hill). This opinion-based probability is expressed in the form of mean and variance (to account for uncertainty). These probabilities are used as priors to create a state transition matrix (specifying the probability of the subject transitioning from each state to all other states) from which to generate a predictive probability distribution map. As a person can only travel to adjacent cells, all of the cells that are not adjacent to the current cell have probability zero. Of course, as the person moves, new cells will become adjacent. Assuming a process in which only present states affect the probability of future states (i.e. assuming a first-order Markov process), it is then possible to predict the lost person’s trajectory as time progresses (i.e. the continuous posterior beliefs of the lost person’s position at different times). While in the ring model the terrain is divided in concentrical rings, in Lin and Goodrich’s model the whole terrain gets divided into tessellated hexagons. In their hypothetical scenario, these hexagons are 24m wide. Each state in the transition matrix is a hexagonal cell in the tessellation. Each row in the matrix is composed of the transitional probabilities of transitioning from one cell to each cell in the tessellation (and of staying in the same cell). Given a point at which the person was last seen (in which the probability of the person being in that cell at the time when she was last seen is 1), it is possible to compute the probability of the person being in different cells at different points in time, thus generating a probability distribution map to guide WiSAR efforts. One advantage of Bayesian models is that existing observations (e.g. GPS track logs and their associated terrain) can be used to update prior beliefs and thus reduce experts’ uncertainty. It is impossible to determine the continuous posterior beliefs about the lost person’s position in closed form, so Lin and Goodrich employ a Markov Chain Monte Carlo (MCMC) algorithm to approximate the posterior distribution needed to generate the updated probability distribution map (following the methods of Gelman et al. 2004). The MCMC method consists on a massive iteration of different transformations from prior distributions (based on expert opinion) to observations (based on GPS track logs) to generate a series of distributions (21 in their hypothetical scenario) that approximate the posterior distributions (see Lin and Goodrich 2010 for detail). These posterior distributions are used in the place of the original prior distributions to update the probability distribution map. The table below sums up some of the distinctions between Koester´s ring model and Lin and Goodrich’s Bayesian model:
Model Sources Updating Terrain sub-division
Ring Model Actual cases (ISRID) Based on time Concentric Circles
— 21 —
Bayesian Model Expert Opinion Based on both time and observations
Hexagonal Tesselation
What is used to generate the probability map in ring models are chiefly the statistics of different subject categories. For example, what determines the distribution of the probability map is not whether a subject in a given category is perceived to be more likely to give preference to a given reorientation strategy, but simply how far from the IPP subjects in the same category were found in previous WiSAR cases. And yet, different ways of conceptualising disorientation still play a role in WiSAR operations that use the ring model. WiSAR teams use their beliefs about lost person behaviour (based on a certain conceptualisation of disorientation) to prioritise the search within the sub-areas of a given probability ring. Moreover, conceptual work can improve the ring model itself is by determining which subject categories should be used for generating the probability map. From a unified theory of disorientation one can hypothesise the existence of new categories, and then use GIS to see if these categories are statistically relevant. Furthermore, a good conceptualisation of disorientation is expected to generate predictions as to what parameters (e.g. age, terrain…) might be statistically relevant for determining new subject categories. The importance of implicit theories of disorientation is clearer when it comes to Lin and Goodrich’s Bayesian model. In their model, expert opinion is used to generate a prior distribution regarding the transitions between three dimensions of terrain (vegetation, slope and topography). This expert opinion is based on a theory of disorientation, be it explicit or implicit, scientific or naive. What is promising is that their method offers a way to incorporate different sources for the generation of a prior distribution, which opens the door to using a unified theory of disorientation as a source for the model, or, alternatively, the opinion of an expert informed by such a theory. Thus, an adequate conceptualisation of disorientation, together with Lin and Goodrich’s Bayesian model offers a way in which developments coming from a broad range of fields can be exploited in order to improve GIS-informed WiSAR operations. With a working theory of disorientation, knowledge about the phenomenon could then be used for establishing the prior distribution that determines the initial probability map. What is more, such distribution could be updated with GIS observation, by using observations such as GPS logs to generate a posterior distribution. That posterior distribution could then be used to discover elements of disorientation that have been overlooked by the overall theory, thus highlighting ways in which the theory can be improved.
IV — CONCLUSION I opened this chapter by reviewing the progress in different fields related to disorientation, such as lost person psychology and the neuroscience of spatial cognition. I then outlined several challenges to unite these separate fields (i.e. differences in levels of analysis and differences in settings) to show that conceptual work was needed to connect these fields. I defended the idea that an adequate theory of disorientation would permit us to use the findings in one field to produce hypotheses in a different field. Afterwards, I reviewed two different methods of WiSAR (a ring model and a Bayesian model) to then analyse how an adequate theory of disorientation could help WiSAR profit from the developments in a broad range of research areas. Importantly, in section 1.2, I argued that a working characterization of disorientation is a necessary first step in the direction of a unified theory. The difficulty is that such a characterization should both spring from the subjective experience of disorientation and be constituted of non-subjective elements. Accordingly, chapters 2 and 3 explore the subjective experience of disorientation, and chapter 4 advances a characterisation of disorientation in non-subjective terms.
— 22 —
2 . T H E P H E N O M E N O L O G Y O F D I S O R I E N T A T I O N
I – SELF-LOCATION AND SELF-CONSCIOUSNESS As I write, my laptop is in front of me, my chair below me, the window to my left, and beyond that, is the city. All of these elements are arranged spatially in my perception, in relation to each other and also in relation to myself, so that I can always locate myself with respect to them. If walking one day without paying much attention I arrived at a familiar square but from an unexpected angle, I might become disoriented. I would still be able to say that I am in such and such square, and the pavement, the road and the surrounding streetlights would still be spatially arranged around me in my perception. Nevertheless, something would have gone awry in my self-location, and there would be a sense in which I don’t fully know where I am. This diminishment in my self-location will in turn change how I experience the space around me, and even how I experience myself within that space. In the present chapter, I will use the relationship between self-location and self-consciousness as a framework to explore the phenomenology of disorientation. Much of the discussion on the relation between orientation and self-consciousness has revolved around the issue of self-location. In his book Varieties of Reference, Gareth Evans argues that self-location is a necessary element of self-consciousness, not just due to the spatiality of the world surrounding the subject, but due to the subject itself being a spatial entity within this world (Evans 1982). Evans has in mind a somehow intellectualised notion of self-consciousness that implies the subject grasping a simple theory explaining how the world results in her perceptions of the world. Jose Bermudez extrapolates Evans’ general idea and brings it to bear on a primitive, non-conceptual notion of self-consciousness. This re-conception emphasizes the active rather than the passive side of self-location: A vital element in a self-conscious subject’s grasp of how he himself is a part of the physical world is indeed derived from an understanding of how his perceptions are a function of his location in precisely the way that Evans brings out. But it must not be forgotten that a vital role in this is played by the subject’s own actions and movement. Appreciating the spatiality of the environment and one’s place in it is largely a function of grasping one’s possibilities for action within that environment. -p.223, Bermudez 1998 In a recent paper, Adrian Alsmith reviews the philosophical literature on self-location and draws a distinction between perspectival and agential accounts of self-location (Alsmith 2017). Perspectival accounts hold that if an experience is perspectivally structured, then it is self-locating (for proponents, see Evans 1982, Cassam 1997 and Schwenkler 2014). In contrast, agentive accounts hold that an experience is self-locating if it represents an object as the focus of its subject’s possible
— 23 —
action. Alsmith highlights Bill Brewer (1992) and Susana Schellenberg (2007) as two major proponents of agentive accounts of self-location. In what follows, I will present four issues with the perspectival self-location account. The first issue concerns photos and films, which have a perspectival structure but do not serve to self-locate the subject with respect to the object of perception. If I am looking at a photo of the Eiffel tower, I am none the wiser as to where I am located with respect to the Eiffel tower: The visual process of looking at photographs fails to carry egocentric spatial information about their depicta. For there is no probabilistic relationship between the photographic image and the egocentric location of the depictum: as I move around the world with the photograph, the egocentric location of the depictum changes, but the photographic image does not. -Cohen and Meskin 2004:201 The same worry applies rather seamlessly to film: “Film experience thus seems to be a candidate case in which one can experience something from elsewhere, without representing oneself as being in that location.” (Alsmith 2017:269). The bottom line is that this type of perceptual experiences serves as a counterexample to the perspectival self-location thesis. These experiences have a perspectival structure, but they do not represent the subject’s location. The second issue comes from experimental studies on whole-body illusions. Let us look at Bigna Lenggenhager and colleagues’ seminal study (2007): Participants wore a VR headset, and the backs of their bodies were filmed and projected as if they were two meters ahead of them. Their backs were then stroked, and participants’ self-location showed a drift towards the virtual body. Huang and colleagues (2017) replicated the study with additional experiments, manipulating the origin of the first-person perspective (e.g., moving the camera progressively away from the subject) in order to study the relationship between bodily self-location and the first-person perspective of visual experience. The resulting changes in both bodily self-location and the perceived origin of the first-person perspective led the researchers to conclude that the two are interrelated but distinct factors. These results sit uncomfortably with perspectival accounts, because they show that the perspectival and the self-location aspects of experience are doubly dissociable. Self-location is not fully specified just by experiencing perspective. Put plainly, in ordinary non-experimental contexts, perspective and self-location are two aspects of the same experience, but they do not seem to be two sides of the same coin. The third issue is that neuropsychological studies show that self-motion information is needed for self-location (MacNeilage et al. 2012; Lopez et al. 2013; Barry and Burgess 2014). This information comes from proprioceptive, vestibular and motor systems. Although optical flow also carries self-motion information, animals do not need visual information to self-locate. For example, gerbils can complete a search task and return to the nest in the dark (Mittelstaedt and Mittelstaedt 1980). And when experimenters introduced error into the animal’s representation of direction by rotating them (slowly, so that their vestibular system did not register it), the gerbils headed back to the nest with an angular error that equalled the angle by which they were rotated. While the above experiment shows that perspective is not necessary for self-location, more recent neuroscientific studies employing virtual environments suggest that it is not sufficient either. Guifen Chen and colleagues (2012) got mice to navigate on a virtual linear track with no self-motion information and in a similar real environment. They observed the firing of the mice’s hippocampal place cells, which provide the neural representation of environmental location, and what they found is that self-motion information was required for normally localized firing in the majority of the place cells, which indicates that visual information alone is not sufficient for ordinary self-location. Further trials in which visual and self-motion information were in conflict resulted in a non-linear
— 24 —
combination of the two to influence place-cell firing. This is consistent with studies of human navigation, according to which we use both self-motion and environmental information in a reliability-weighted Bayes-optimal fashion (Nardini et al 2008). This leads us to the fourth and final issue, which concerns the point (or lack thereof) at which we are located in our experience. A perspective has a single point of origin (e.g. in between our eyes in the case of visual experience). However, in our ordinary self-locating experience, “there is no privileged point in (or on) my body that counts as me for purposes of characterizing my perceived spatial relation to the object” (Briscoe, 2009:425). I don’t experience myself as being located as the point of origin of the perspectival structure of visual experience nor at any other single point (say, either of my ears if we considered auditory self-location). The issue goes further because we do not experience ourselves as multiply located either. We experience ourselves as a bodily self that occupies and is located in an extended region of space. Alsmith (2017) argues that there is a critical difference between locating perspectives and locating subjects. Different perceptual systems representing space locate the origin of perception as different single points in space (e.g. the eyes in the case of visual perception and the ears in the case of auditory perception). Thus, it would follow from perspectival accounts of self-location that subjects are multiply located, while subjects seem to self-locate in a unitary way. Christopher Peacock makes a similar point when he says that “a point of view in space cannot be the producer” of an action (p. 133, Peacock 2019). According to Alsmith, while perspectival accounts fail to account for the unitary character of self-location, agentive accounts of self-location are well-equipped to deal with this problem:
The issue is easily resolved by properly appreciating the significance of the fact that the object is represented as the focus of the subject’s possible action. To represent an object in this way is not to represent it in relation to a single position from which one perceives and acts. Rather it is to represent it in relation to a single, unified thing from which one perceives and with which one acts.
-p.278, Alsmith 2017 Given the four issues with perspectival accounts presented above, I will endorse an agentive understanding of self-location in the phenomenological analysis that follows. Before diving into the phenomenology of disorientation, it is important to discuss a point of disagreement between Brewer and Schellenberg, two of the major proponents of agentive accounts of self-location. The disagreement concerns the role that egocentric (self-referenced) and allocentric (world-referenced) frames of refence play in self-location. Egocentric reference frames are coded in relation to the subject’s body, position and orientation. Allocentric refence frames are coded independently of the subject, based on cues, geometrical features or cardinal directions (Klatzky 1998). While Brewer’s focus is on a primitive perceptual egocentricity that represents perceptual objects in spatial relation to the subject, which he claims doesn’t require a system of allocentric representations (Brewer 1992), Schellenberg thinks that self-location requires the subject moving from egocentric to allocentric frames of reference1. In particular, it requires what she deems
1 In some of his recent work, Bermudez defends the view that allocentric frames of reference are necessary to be
conceptually self-conscious, but that only egocentric frames of reference are necessary to be non-conceptually self-conscious: “To be non-conceptually self-conscious is to be in receipt of and be able to act upon, locally self-specifying information—that is, information about one’s immediate bodily disposition and about one’s spatial relation to the local environment. Nonconceptual self-consciousness is paradigmatically underpinned by somatic proprioception and by the outward-directed senses. Conceptual self-consciousness brings a degree of emancipation from that perception-dependence. It allows the thinker to act and form plans in ways that are not bound either by the present time or by the present location” (p.113, Bermudez 2017). In what follows, I simply hope to show that
— 25 —
entertaining alter-ego points of view on the object. An alter-ego point of view is an engaged allocentric frame of refence, engaged in the sense that it is “the focal point of a potential egocentric frame of reference” (p. 463, Schellenberg 2007). There are two objections that I would like to raise against Schellenberg’s view. The first one is that all allocentric frames of reference are focal points of potential egocentric frames of reference, because there is a linearly transformable relationship between egocentric and allocentric coordinates (Ekstrom et al 2017). There is no way to make sense of alter-ego points of view being allocentric frame of reference that are the focal points of a potential egocentric frame of reference, because all allocentric frames of reference fulfil that description, making it impossible to distinguish engaged from non-engaged allocentric frames of reference. A viable alternative is to understand the distinction between engaged and disengaged allocentric frames of reference as a distinction between online and offline systems of spatial representation. To understand this online-offline distinction, we have to look at the disorientation studies of Wang and Spelke and of Waller and Hodgson. Wang and Spelke trained subjects to learn the array of objects in a room and then point to unseen targets. They compared the performance of oriented and disoriented (through self-rotation) subjects and found that disorientation impaired both the subject’s relative accuracy in pointing to different objects and their absolute accuracy in pointing to all objects (Wang and Spelke 2000). Waller and Hodgson then extended the above experimental design by asking subjects to judge the relative directions between different objects in the room (e.g. “imagine that you are at the door, facing the couch, point to the TV”) and found that disorientation actually increases inter-object pointing accuracy, supporting the idea that disorientation elicits a switch from an online system of spatial representation to an offline system of spatial representation (Waller and Hodgson 2006). While the offline system produces enduring and stable spatial representations of the environment, the online system produces representations that are transient and dynamically updated through spatial perception, which makes the online system relatively more precise (i.e. less coarse) than the offline system, but unstable and unprecise during disorientation. According to Waller and Hodgson, if the online representation system was active, it was harder for the subjects to focus on a task (inter-object spatial relations) that required only an offline spatial representation. Thus, the task was easier for disoriented individuals that were operating based solely on an offline system of spatial representation, resulting in their increased performance. To go back to the discussion of Schellenberg’s notion of engaged allocentric frames of reference, we should clarify that in normal circumstances, the online system of spatial representation integrates both allocentric and egocentric frames of reference in a dynamic way. In this sense, one can say that the relevant allocentric frames of reference are engaged as long as the online system of spatial representation is active. My second, more substantial objection to Schellenberg’s view is that it is unclear that an allocentric frame of reference is actually necessary for self-location. Take patient CF, who has an impairment locating objects relative to her environment but not relative to the perceived viewpoint. Burgess and colleagues ran a series of tests on patient CF that suggest “the presence of a selective deficit of allocentric spatial memory” (p.246, Burgess et al 2006), that is, a deficit in using allocentric frames of reference. Importantly, patient CF showed no impediment in using egocentric frames of reference. This impairment in using allocentric frames of reference resulted in severe navigational impediments, but her performance on a broad range of different tests, including space perception tests, was flawless. Patient CF showed “spared recognition of landmarks, object locations and spatial scenes from the same point of view, and unimpaired spatial abilities such as mental rotation” (p.248, Burgess et al 2006).
allocentric frames of reference are not necessary for self-location, without delving into the distinction between conceptual and non-conceptual self-consciousness.
— 26 —
I raise the case of patient CF case not necessarily as an objection against the core of Schellenberg’s agential account of self-location, but only against the role that allocentric frames of reference play in arguing for that account. At heart, what Schellenberg’s account seems to require is not that the subject knows precisely how the environment would look from a different location, but just that she has the capacity to entertain the possibility of re-locating. After all, what Schellenberg defends as the core of her thesis on spatial perception is that what makes spatial perception possible is not action tokens but the capacity to act, so that “perceiving intrinsic spatial properties requires perceiving objects as perceivable from locations other than the one that one happens to occupy” (p. 661, Schellenberg 2007). In a similar way, what makes self-location possible is not tokens of mental re-location (i.e. engaged allocentric frames of reference), but the capacity to re-locate. An allocentric frame of reference might be needed for the subject to know what an environment might look like from a different viewpoint, but it is not needed for the subject to tacitly know that moving within this environment will result in a new viewpoint and that she has the capacity necessary to bring about this movement. While patient CF doesn’t have the knowledge of what an environment might look like from a different viewpoint, she still has the tacit knowledge required for spatial perception. Following this line of thought, it seems that while egocentric frames of reference might be necessary for self-consciousness, allocentric frames of reference are definitely not. This points to a difference between two notions of self-location, which I will term minimal self-location and integrated self-location. Minimal self-location involves only egocentric frames of reference. Integrated self-location, on the other hand, involves the integration of egocentric and allocentric frames of reference through the online system of spatial representation. Integrated self-location is the norm in everyday human activity, and it underpins our standard navigational abilities. When this capacity for integrated self-location is compromised during disorientation, disoriented individuals can still orient in a minimal sense (extreme cases like immersion in sensory deprivation tanks being a possible exception outside the scope of this chapter). The same is true of patients with navigational impairments such as CF, who can still perceive the spatial arrangement of objects around her and act accordingly. Similarly, a disoriented individual that cannot properly navigate within her environment, is still able to perceive that a given door is to her left, and she can walk to that door or reach out to grab a cup of coffee on the table to her right, because she is still able to self-locate minimally. Here, I will remain neutral on the question of whether or not minimal self-location is required for self-consciousness2. My claim at this point is a negative one: integrated self-location (which is the standard case of self-location in non-pathological cases) is not necessary for self-consciousness. A large part of the philosophical discussion of self-consciousness revolves around the capacity for first-person thought, but even if we grant that there is a form of self-consciousness in experience that grounds the capacity to entertain first-person thought, the study of first-person thought does not exhaust the study of self-consciousness. My focus in this chapter will not be on what forms self-consciousness are necessary for first-person thought or on what capacities might be required for self-consciousness. Rather, I am interested in exploring the form of self-consciousness that is characteristic of our ordinary oriented experience of the world and in how this form of self-consciousness is transformed during episodes of spatial disorientation. Consequently, one of the central aims of this chapter is to investigate the role that integrated self-location plays in shaping our everyday self-consciousness, even if, as I have just argued, self-consciousness can occur without integrated self-location. 2 The widely held view that (minimal, following my distinction) self-location is necessary for self-consciousness is not
without its opponents. Raphaël Millière appeals to the case of a deafblind individual (MVS) to defend that self-location is not necessary for self-consciousness (Millière 2019).
— 27 —
In what follows, I will use the distinction between minimal and integrated self-location to assist in a phenomenological investigation of spatial disorientation.
II– DISORIENTATION ALONG THE THREE EGOCENTRIC AXES
Suppose you were in some place that you know perfectly well, such as your study, or the street in which you live; suppose that you shut your eyes for a second, and that in this short interval the external world about you turns round to the extent of two right angles on a vertical axis passing through your body, so that an object previously in front of you should be behind you, another placed on your right hand should have shifted to your left, and so on; suppose that in this rotation the external world should be displaced as a whole, and that the objects, notwithstanding their change of position in regard to you, should retain exactly their relations to one another; lastly, suppose that during this strange revolution you have not moved, but that, on the contrary, you have the firm conviction that you have been absolutely at rest. Now open your eyes and look around you, and fancy the feeling of bewilderment that would seize you, and you will have an idea of the impression I experienced when I was under the attack of what is called vertigo of direction
-report in p. 156, Binet, 1885
The feeling of being turned around was first mentioned in the context of a discussion on instinct started by Charles Darwin in 1873 in the journal Nature (see Schmidt di Friedberg 2018 for a review) and was first explored methodically in Alfred Binet’s articles “Vertigo of Direction” and “Reverse Illusions of Orientation” (Binet 1885, 1894). Binet’s work is built primarily on reports from subjects who had been turned around, a state that Binet initially termed vertigo of direction3. Subjects report an illusion of orientation, sometimes even characterised as a type of hallucination, which is usually accompanied by distress and anxiety. The illusion carries the impression that the space around the subject is aligned in the opposite direction than it should, so that it seems to the subjects as if they have been unknowingly turned around. One subject reports:
My confusion was extreme, and I found myself for some time in a state of complete disorientation, and unable to comprehend the situation. [...] All the objects and streets occupy positions exactly opposite to those which they should.
-report from p. 340, Binet 1894 As one of the subjects emphasizes, being turned around is very different to finding oneself lost in an unknown place or visiting a new environment:
Here you know very well where you are. You have a very clear sense of direction and you know perfectly where things ought to be; only this direction and this place are just the opposite of their real position. The station of Mont Parnassus ought to be on the right, and I did not understand why I did not see it there.
-report from p. 340, ibid
3 The sense in which being “turned around” is similar to vertigo is that during both experiences the world around one
appears to hover and only “clicks back into place” when the feeling recedes. However, as Binet himself concedes in his later work, the term “vertigo of direction” is inexact and rather misleading. Therefore, I will not be using this term in the present chapter.
— 28 —
Being turned around is quite a common occurrence. As for myself, I often get turned around when I come out of a metro station and I take the exit on the wrong side of the street by mistake, so that what I expect to find before me appears behind me and vice versa. In these cases, once I figure out my mistake, the illusion fades away, as if the environment had again turned in the right direction. Subjects are often bewildered or perplexed after being turned around, and sometimes they are even unable to recognise familiar environments, as in the following case:
Instead of taking the right to return to the Place de la Republique, I took the left toward the Hotel de Ville…While on my way I felt sure of meeting the Place de la Republique. Thus my confusion was extreme on coming to the Hôtel de Ville… I was some moments in recognizing it. Then I recognized the Hôtel de Ville, without destroying the illusion. It disappeared, however, very quickly… when I understood the cause of my mistake.
-report from p. 341, ibid Following the discussion in the previous section, we can redescribe the phenomenon of being turned around as a misalignment of the subject’s allocentric and egocentric frames of reference, caused by errors in the online system of spatial representation and resulting in a degraded integrated self-location. The exact cause of the errors is outside of the scope of this chapter and it might vary on a case by case basis. In some cases, it is the subject’s misguided conscious inferences or navigational decisions that result in a mistaken integration. In others, it might be due to an accumulation of error in sub-processes that contribute to the online system of spatial representation (e.g. path integration) or the result of biases in these sub-processes. There are at least four aspects of the experience of being turned around that are worth discussing in the context of self-location and self-consciousness:
1) it is associated with feelings of distress and anxiety; 2) it causes perplexity, unfamiliarity and often prevents recognition of well-known
environments; 3) egocentric frames of reference are not affected by it; and lastly, 4) the allocentric frames of reference are almost always misaligned 180º.
We will go back to the associated feelings that come with disorientation experiences (not solely those of being turned around) in section 2.3, and in this section, we will discuss the three remaining aspects. The failure in environment recognition and the perdurance of egocentric frames of reference are best understood if studied together. Let us take the case of one of Binet’s subjects who declares that the landscape appeared as if it had turned itself around. Clearly, what he means is not that he saw the environment around him literally turning or that instead of seeing what was in front of his torso he was seeing what was behind the back of his head. And in the case of the subject who unexpectedly comes upon Hôtel de Ville, he sees the building, but he cannot recognise it. The sense in which the landscape “turns around” (or analogously, the sense in which the subject gets turned around within the landscape) is that objects are in the opposite direction of where they are expected to be based on an allocentric frame of reference that is off by 180º. Hence the perplexity and the bewilderment. Nevertheless, an egocentric frame of reference remains, and this suffices to maintain correct minimal self-location. That is, the turned around subject is still able to indirectly locate herself with respect to objects present in her visual field. She perceives that a sycamore tree is to her right, that some children are shouting somewhere behind her or that such and such a building are there, at a given distance from her body and at given angle from the direction she is facing. Binet’s subject still sees a big building in front of him, he is only unable to identify that building as Hôtel
— 29 —
de Ville, because of the misalignment between his allocentric and egocentric frames of reference. Importantly, in this case it is not that the integrated self-location vanishes, so that only minimal self-location remains (as in the case of patient CF). The integrated self-location is still there, only in a more uncertain form and associated with the illusory feeling of being turned around. What being turned around reveals is that even if we granted that only egocentric frames of reference are required for minimal self-location and thus for self-consciousness, allocentric frames of reference still play an important role in shaping self-consciousness. For instance, allocentric frames of reference support recognition, which is made evident by the occasional failure in recognising familiar environments after being turned around. And even in those cases in which the turned around subject recognises the places around her (e.g. I recognise station Montparnasse, which is to my right, but it feels like it should be to my left) a feeling of unfamiliarity usually emerges that changes how the environment is perceived and how the subject feels towards that unfamiliar environment. Finally, a turned around subject self-locates minimally with respect to such and such a building (e.g. Hôtel de Ville), but fails to self-locate with respect to a larger environment (e.g. the spatial configuration of Paris), and as we will see in section 3, this has a substantial effect on her self-consciousness. The question remains why the allocentric and egocentric frames of reference should be misaligned by precisely 180º, as it happens in the majority of the reported cases of being turned around. Here, it is necessary to introduce the three axes of egocentric space: frontward-backward, left-right and up-down. According to John Campbell, what characterises these axes is that the subject uses them immediately for action by exploiting ways in which humans are not symmetrical: “If we were symmetrical, being double-jointed and able to look either way, then our current notions of in front and behind simply could not be applied to ourselves, could not guide our actions in the way that they do.” (p.15, Campbell 1995). While it is easy to see how this applies to the frontward-backward axis, it becomes harder when it comes to the other two axes. As Campbell himself points out, up-down has an immediate relevance for action but is better characterised as dependent on the gravitational field rather than on the longitudinal axis of the body. Moreover, there is no notable left-right asymmetry that could be employed for action in the way that Campbell suggests. Nevertheless, something that makes the left-right distinction paramount is the role it plays in spatial orientation. This is a point that Immanuel Kant illustrated quite well in his 1768 essay “What Does it Mean to Orient Oneself in Thinking”:
In the proper meaning of the word, to orient oneself means to use a given direction (when we divide the horizon into four of them) in order to find the others - literally, to find the sunrise. Now if I see the sun in the sky and know it is now midday, then I know how to find south, west, north, and east. For this, however, I also need the feeling of a difference in my own subject, namely, the difference between my right and left hands. I call this a feeling because these two sides outwardly display no designatable difference in intuition. If I did not have this faculty of distinguishing, without the need of any difference in the objects, between moving from left to right and right to left and moving in the opposite direction and there by determining a priori a difference in the position of the objects, then in describing a circle I would not know whether west was right or left of the southernmost point of the horizon, or whether I should complete the circle by moving north and east and thus back to south. Thus, even with all the objective data of the sky, I orient myself geographically only through a subjective ground of differentiation.
-p.7-8, Kant 1768 This subjective ground of differentiation between left and right is given ostensibly in a non-conceptual form (De Monticelli, 1986). However, the left-right axis is necessary but not sufficient for the kind of orientation that Kant has in mind — in a three-dimensional space, three egocentric
— 30 —
axes are required. This becomes clearer when we reconsider the case in terms of integrated self-location. What Kant is describing is integrating an allocentric frame of reference (the four cardinal directions) with an egocentric frame of reference based on a single cardinal direction that is present in both (i.e. the sun at noon). When we are facing North, West is to our left, East is to our right and South is backwards. However, this requires that we face north, and that we do so without being unknowingly upside down. If we were in a space in which we couldn’t tell whether we were upside down or not (e.g. a zero-gravity space that was symmetrical along the vertical axis), knowing that we are facing north would permit us to locate south (which would be behind us), but not to locate east and west. Conversely, if we were not able to differentiate backward from forward, we would be in a similar position to a creature unable to tell left from right through a subjective ground of differentiation. The reason that this latter point is often overlooked is that while the left-right distinction requires a subjective ground of differentiation (and one that we need to learn and that humans lack in early infancy), as Campbell indicates, the difference between backward and forward is given through the asymmetry of our own body along the forward-backward axis. To further support the idea that the left-right axis (unlike the up-down and forward-backward axes) does not play a significant role in driving action, I will look at some of the research on defensive peripersonal space (DPPS). Animals respond to stimulus more strongly when the stimulus takes place in close proximity to the body, within the DPPS region (Graziano and Cooke 2005). A recent study mapped the face’s DPPS and found asymmetries in its distribution: the face’s DPPS is projected forward (as opposed to backward) and upward (Bufacchi and Iannetti 2016). In line with Campbell’s view, while the forward-backward asymmetry depends on the bodily frontward-backward axis (so that if the subject lies down the DPPS is still larger before the subject’s head than behind it), the up-down asymmetry depends on gravity and not on the subject’s body (so that if the subject lies down, looking up or looking sideways, the DPPS becomes larger towards the ceiling, not towards the “top” of the subject’s longitudinal axis). The reason to bring this topic into our discussion is that the experimenters found no left-right asymmetry in the geometry of DPPS, which supports the idea that while the up-down and frontward-backward axes have immediate relevance for action, this is not the case for the left-right axis4. If this is correct, then it seems that one of the fundamental roles of the left-right axis is integrated self-location, which would not be possible without this axis. That all three egocentric axes are necessary for integrated self-location gains further support when we discover that illusions of orientation do not happen only along the frontward-backward axis, but also along the left-right axis and even along the up-down axis. Flournoy, in his work Des Phénomènes de Synopsie, describes a left-right axis inversion while waking in the dark at night:
We clearly know that we are lying in bed with the right side toward the wall, and in spite of that we feel that we have the wall to the left and the room to the right and are flustered at the tenacity of this illusion, which can resist reasoning for many seconds; unless, extending the arms, the contact with the wall causes it to vanish suddenly and brings the mental images back into place.
-p.188, Flournoy 1893, my translation The difficulty here is that most of our allocentric frames of reference are already oriented along both the frontward-backward and the left-right axis, so that a 180º misalignment implies a misalignment with respect to both axes. Flournoy’s description could be taken as a left-right inversion or as a 180º reversal (inversely, some of the Binet cases could be taken as a left-right
4 A further review of studies about different peripersonal spaces (i.e. not only DPPS) does not report any geometric
asymmetry along the left-right axis either (Bufacchi and Iannetti 2018). While it is conceivable that in some cases a left-right asymmetry could emerge (e.g. for tennis players who are not ambidextrous) there is no empirical support for this, and it would be the exception rather than the rule.
— 31 —
inversion). A better example would be the following thought experiment: Imagine a familiar rectangular room in your office that has no windows and a single door in the middle of one of its walls. One night, as a prank, some co-workers change the spatial arrangement of the objects within the room in a mirror-way with respect to an axis of symmetry drawn perpendicularly from the door. Now imagine that you enter the room unaware of the change. The objects that you would expect to your right would be to your left and vice versa, producing a left-right inversion illusion. The same effect could be achieved using a virtual reality environment. For the sake of argument, I will take Flournoy’s anecdote as a case of left-right inversion. Something that I find specially interesting about Flournoy’s report is that, as it happens in darkness, the illusion is not primarily visual. This is also the case in the 180º reversals mentioned earlier, but what the darkness element in Flournoy’s story makes particularly noticeable is that the experience of integrated self-location is transmodal. Flournoy senses that the wall is to his left, but the wall is nowhere to be seen on either his left or his right. The wall is expected to be on his left based on the (erroneous) integration of allocentric and egocentric frames of reference and this integrated self-location generates the expectation that were he to turn on the light, he would see the wall to his left and that were he to reach to his left he would touch the wall. Finding the wall to his right, consequently, makes the illusion vanish. The transmodality of integrated self-location means that it has an influence on both our perception (as we saw in Binet’s reported cases of non-recognition and unfamiliarity) and our experienced possibilities for action (e.g. reaching to the left to touch the wall), which is what one would expect following an agentive account of self-location. It seems that if cases of left-right inversion are hard to find (we need darkness, thought experiments or virtual reality), cases of upside-down reversal would simply not exist. For how could one not know if one is upside-down or not? Not only is our world clearly different along the vertical axis, with furniture and carpets on the ground and lamps on the ceiling, but gravity always tells us where down is. Even if someone had stuck furniture to the ceiling as a prank, we wouldn’t be fooled. We would simply feel that the furniture is upside down, without undergoing any subsequent illusion. Luckily for us, up-down reversals exist in micro-gravity settings, as reported by Skylab (Cooper, 1976; Johnston and Dietlein, 1977) and Spacelab astronauts5 (Oman et al. 1986, 87, 90):
Two of the most common situations were when working upside down (relative to their normal 1-G orientation in training), or when floating right side up but viewing another crewmember floating upside down in the cabin. In either case, crew often experienced the striking illusion that the surrounding walls, ceiling, and floors had somehow exchanged identities. In the first situation, whichever surface was closest to their feet seemed like a generic floor. Surfaces approximately parallel to their body now seemed like walls, and overhead surfaces were perceived as ceilings. In the second situation, the orientation of the inverted crewmember determined the direction to the “floor” … Areas of Skylab that had locally incongruent visual vertical cues also triggered VRIs [Visual Reorientation Illusions].
-p. 211-212, Oman 2007
These illusions of orientation along the up-down axis show similar characteristics associated to being turned around illusions such as perplexity, lack of recognition, unfamiliarity and distress. However, as one could have expected, the effects of up-down axis illusions are much stronger, often provoking nausea and even vomiting for the astronauts.
5 Similar experiences have been reported by air pilots and by divers. Note also that some of the environments in which
astronauts show no clear up-down distinctions (e.g. contrast normal rooms in which the direction of the furniture indicates where the floor is with a tubular Skylab corridor with controls both on the ceiling and on the floor).
— 32 —
Together, the illusions along the three egocentric axes point to the role of the body in integrated self-location, in line with Husserl’s notion of the body as the zero point (“Null Punkt”) of orientation (Husserl 2012a):
The body, for Husserl, not only has an orientation in space, it also orients space around it … Every space is experienced from the inescapable ‘here’ of my body: right and left, up and down, near and far. All orientation involves a body and all distances are marked off taking the body as the point of departure.
-p.450, Dermot and Cohen, 2012
And as we saw earlier, for the body to work as the zero point of orientation all three axis are needed, and one of these axes (up-down) depends on gravity, not on the body itself. It is only given the up-down axis delineated by gravity that the body becomes the meeting point of the left-right and frontwards-backwards axes and orients space around itself. While the agentive accounts of self-location introduced in section 2.1 highlight the role of the body in minimal self-location as the necessary origin of one’s capacity to act, Husserl’s “Null Punkt” highlights the role of the body in integrated self-location as the necessary origin of the left-right and frontward-backward axes that connect egocentric and allocentric frames of reference. In this section, we have seen that illusions of orientation can happen along the three egocentric axes, supporting the idea that all three egocentric axes are involved in generating an integrated self-location. And while the left-right axis does not seem to play an immediate role for action, it does seem to play a fundamental role in the alignment of allocentric and egocentric frames of reference. The effects of different illusions of orientation also reveal that integrated self-location is transmodal and therefore has an effect on both our perception (e.g. influencing recognition) and our experienced possibilities for action. Finally, we have started to bring to the foreground some of the feelings often associated with disorientation (such as perplexity, unfamiliarity and distress), which will be explored more in-depth in the next section.
III– THE PHENOMENOLOGY OF DISORIENTATION
The “bell-tower of Marcellinara” has become the symbol of this form of existential anguish and sense of disorientation. De Martino, in La fine del mondo, recounts an experience of his own in 1950s’ Calabria, when he and a friend stopped in the village of Marcellinara to enquire for directions from an elderly shepherd. To preclude any possibility of error, De Martino asked the shepherd to accompany them as far as the turning that they needed to take a few kilometres ahead: “He got into the car somewhat reluctantly, as though he feared some kind of trap, and this reluctance gradually grew into distress, because from the car window, out of which he was constantly peering, he had lost sight of the bell-tower of Marcellinara, the point of reference for his extremely circumscribed domestic space. Due to the disappearance of the bell-tower, the poor old man felt completely disoriented: and only with difficulty did we manage to get him to the right turning and to obtain the information that we needed.” (De Martino, 1977, pp. 480–481) At last they brought the shepherd back to the village. By now he was in a state of great anxiety, and, in his impatience to see the bell-tower reappear, travelled with his head leaning out of the
— 33 —
car window: “When he finally saw it, his face relaxed and his aged heart began to quieten, as though he had won back a ‘lost homeland’” (ibid., 1977, pp. 480–481).
-p. 6, Marcella Schmidt di Friedberg 2018
The anxiety described in the “bell-tower of Marcellinara” scenario is common to many episodes of disorientation, although of course this anxiety can occur with different degrees of intensity. Together with the Disorientation research team at Institut Jean Nicod, I have been collecting reports of disorientation episodes in order to build a corpus of experiences of disorientation that can help us characterize the phenomenon, in the style in which Binet used subjective reports to characterize the experience of being ‘turned around’. The existing reports collected so far in this corpus give us some very valuable information about disorientation experiences6. Some of the episodes of disorientation reported are illusions along the three egocentric axes (mostly along the frontward-backward axis, although there is one case of an up-down inversion while scuba diving), but these are not the majority. There are cases that involve exploring new environments (e.g. New York subway), cases in which not remembering well what one expected to be a familiar environment prompts disorientation (“In a mall, I disoriented myself trying to find a specific store. I had a vague memory of the place and walked in that direction. I started to feel disoriented when I didn't remember the stores I was seeing”), and cases in which subjects are unable to interpret a map (e.g. the Tokyo Subway map) or in which their GPS stops working (“I was in New York City, and I was trying to figure out if I should walk to the left or to the right after exiting a restaurant. The mobile GPS was not calibrating, and so no matter the direction I walked in, the GPS pointed to the opposite direction”). One of the main benefits of the corpus is that it reveals a series of feelings that are often reported in relation with disorientation and that can help us understand the phenomenon:
• anxiety (“I didn't know where I was heading nor to which side of the station I was going. The feeling is horrible and stressful”; “I was anxious really anxious because I didn't want to be late to my first class”)
• vulnerability or helplessness (“it made me feel vulnerable”; “I felt unsafe and
anxious because no one was with me, and the environment was not familiar to me”; “I just felt confused and helpless — I didn’t know what to do”)
• confusion or perplexity (“I was slightly confused, as all of the streets seemed similar
and we kept walking in a circle although we took different turns”)
• diminishment or isolation (“I get nervous the longer it takes me to figure out the direction. I attempt to find out which street/avenue I am at. It feels like I am a tiny spec in all this action that’s happening around me”; “Feeling of being left alone, feeling of not being able to make decisions and unable to be independent.”).
This is in line with the feelings that subjects report in Binet’s corpus. Furthermore, 66% of subjects agreed (from somewhat agree to strongly agree in a Likert scale) that the experience of disorientation made the environment feel unfamiliar, although no subject in our survey explicitly reported lack of recognition of the kind that Binet’s subjects occasionally talk about. It might well be that lack of recognition is mainly due to misalignments along an egocentric axis, as this is an effect that astronauts also report experiencing when they undergo an up-down inversion illusion.
6 As of 29 January 2021, we have received 192 responses.
— 34 —
Something that our corpus makes particularly clear is the presence of feelings of vulnerability, helplessness and diminishment or isolation that emerge in relation with disorientation. These feelings are particularly worth reflecting on in the context of self-consciousness. In what follows, I will pursue a phenomenological analysis in an effort to better understand the vulnerability, helplessness, diminishment or isolation that often arrive when experiencing disorientation.7 One of the ways in which anxiety manifests during disorientation is through pressure on one’s chest — a restriction that goes hand in hand with the sense of diminishment that many subjects report8. Here, it is illuminating to look at Thomas Fuchs’s work on the phenomenology of depression for an example of the connection between anxiety and diminishment. During depression there is a diffuse anxiety and an overall bodily rigidity, so that “the external aims and objects, as it were, withdraw from the patient; they are not at the disposal of his body as a matter of course … All this means that the body’s space shrinks to the nearest environment” (p. 110, Fuchs, 2005). When talking about the body’s space, Fuchs is referring to the body as the medium of our embodied experience of being-in-the-world, the lived-body. Fuchs’ characterisation of anxiety in depression is similar to Heidegger’s view that “in anxiety, what is environmentally ready-at-hand sinks away” (p.232, Heidegger 1962). The idea of body-space reduction (hereon BSR) is very fruitful for understanding the phenomenology of disorientation, and it resonates with the subjective reports of disorientation experiences and the oft-reported feelings of diminishment, vulnerability and helplessness (e.g. “it feels like I am a tiny spec in all this action that’s happening around me”). The point is not that these three feelings can be equated to a single feeling (namely, the BSR that is characteristic of anxiety) or even that they must necessarily occur together, but simply that all of these feelings are strongly associated and integrated together during most disorientation experiences, and that this association is well-captured with the concept of BSR. Something that Binet’s subjects are careful to point out is that the anxiety comes as a result of the disorientation and does not precede it. The same can be said of the feeling of unfamiliarity — it is not the cause but the result of the disorientation. This leaves us with the question of what is it in the disorientation experience that brings about both the unfamiliarity and the anxiety (i.e. BSR). After all, anxiety and unfamiliarity alone do not constitute disorientation, and there is something that makes the experience of disorientation unique and different from other experiences involving unfamiliarity or anxiety — different from depression for instance. I have already claimed that during disorientation there is a degradation in the subject’s integrated self-location, but more needs to be said about how this degradation appears in the phenomenology of disorientation itself if we want to understand the connection between disorientation and self-consciousness. Charles Oman compares the micro-gravity VRIs that were discussed in the previous section with figure-ground illusions (such as the profiles/vase illusion): “VRIs typically occur spontaneously, but as with figure ground illusions, onset depends on visual attention and is therefore under cognitive control. One astronaut commented: ‘If you really want a surface to be “down”, you can just look at it and decide that it is’” (p. 213, Olman 2007). This kind of cognitive control can also be exerted in left-right inversions in darkness like the one described by M. Flournoy, and I have noticed myself that even during experiences of being “turned around”, how I allocate my attention influences the persistence or extinction of the illusion, so that I can just let the illusion be or I can make a concentrated effort to mentally rotate my allocentric frame of reference and extinguish the illusion.
7 Like most people, when disoriented, I usually get anxious and quickly try to reorient myself. However, I have recently developed a practice of reflecting on the experience whenever I naturally happen to become disoriented (i.e. without intently looking for said experiences through psychogeographical drifting or similar methods). This series of phenomenological exercises inform to a great extent the following discussion. 8 As Fuchs points out, anxiety comes from the Latin “angustiae” which means “narrows”, “restriction” (Fusch 2005).
— 35 —
What the analogy between illusions of orientation and figure-ground illusions reveals is that just as there is a perceptual instability in figure-ground illusions, there is a different type of instability when it comes to disorientation. During disorientation, the online system of spatial representation fails to integrate egocentric and allocentric frames of reference satisfactorily and this brings about a destabilization in integrated self-location. The way that this transmodal destabilization appears in experience is through an increase in uncertainty of what lies beyond one’s egocentric space. When I become disoriented, it becomes uncertain whether behind a row of buildings is a given part of the city or another. It is as if the rest of the city were hovering behind my field of vision, undecided about its position — in some cases it even feels like the city is no longer there behind the buildings. Another Husserlian concept, “the horizon of experience”, can help us understand how this destabilization in integrated self-location results in BSR and unfamiliarity:
Everything that genuinely appears is an appearing thing only by virtue of being intertwined and permeated with an intentional empty horizon, that is, by virtue of being surrounded by a halo of emptiness with respect to appearance. It is an emptiness that is not a nothingness, but an emptiness to be filled out; it is a determinable indeterminacy.
-p.41, Husserl 2012b The idea is that every experience happens within a framework of references that carries a given set of possibilities, and that this framework is incorporated in the experience itself. The allocentric frames of reference present in integrated self-location would be part of this horizon of experience, carrying with them a set of possibilities, just as the allocentric representation of having the wall to the left in the dark room of Flounoy’s story carried with it a set of possibilities for action (reaching to his left and touching the wall) and for perception (turning on the light and finding the wall to his left). To get an intuitive grasp of the concept of the horizon of experience (particularly in terms of spatial allocentric frames of reference), a simple phenomenological scenario should suffice: While you are in a familiar room, your office for instance, you only see the elements that are directly around you —such as a chair or the door—, but you feel the location of a set of invisible spatial objects —such as the corridor, the main entrance of the building, and some local landmarks surrounding you— that are hovering somewhere out of sight, framing your experience of the space in which you find yourself. These out-of-sight elements are part of your horizon of experience. The horizon of experience is a determinable indeterminancy, which means that it can be determined with varying levels of (un)certainty. The destabilization in integrated self-location that occurs during disorientation is a destabilization of the horizon of experience, which becomes more uncertain as a result. In his later work, Husserl distinguishes the horizons of well-known worlds (homeworld or “Heimwelt”) from those of strange or alien worlds (“Fremdwelt”), so that there are familiar and unfamiliar horizons of experience (Husserl 1970). By destabilizing the horizon of experience, disorientation makes the subject’s world unfamiliar. As we saw earlier in the case of non-recognition after being turned around, an allocentric frame of reference that is part of integrated self-location contributes to perception. Therefore, a phenomenal “homeworld” that has undergone destabilization in integrated self-location will become unfamiliar or strange in lack of this frame of reference — the subject will no longer be at home in the world9.
9 Here, “being at home in the world” simply refers to a subject who is in her homeworld (Heimwelt) as opposed to an
alien world (Fremdwelt). Although this is similar to the use of the term in discussions of Heidegger’s work and although Heidegger himself claims that anxiety is a form of being not-at-home (p. 233, Heidegger 1962), I would not want to frame the discussion in Heideggerian terms. The reason for this is that whether or not the subject can ever be at home in the world in the Heideggerian sense is an open discussion. Dreyfus defends in Hedegger’s view, “Dasein is unheimlich [unsettled], that is, never truly at home in the world” (p.45, Dreyfus 1991), which means that for Heidegger, there is no state of “being at home in the world” that disorientation could destabilise.
— 36 —
Because the horizon of experience carries with it a given set of possibilities, a destabilization of that horizon results in an destabilisation in the experienced possibilities of the subject, so that the possibilities for action become less salient, more ambiguous or simply disappear. A subject that after being turned around doesn’t recognise Hôtel de Ville, for example, perceives that the building in front of him can be entered (just like any other building), but not that the inside of Hôtel de Ville can be reached by entering that building — this is not a possibility that lies in his horizon of experience, and no more is the possibility of going around the building and reaching the Seine open to the disoriented subject. Furthermore, he knows that the building in front of him can be entered, but not that he himself can enter it. It is not only that he cannot know whether or not the building in question is open to the public, but that in the strange world in which he now finds himself, he is no longer “welcome” to explore unfamiliar buildings — the perceived possibility of entering unknown buildings has become weakened as a result of the disorientation. The destabilization that occurs during disorientation shrinks the subject’s possibility space, and this underlies the diminishment, vulnerability and helplessness associated with BSR. The above analysis gives us all of the elements that we need in order to see the effect that disorientation has on self-consciousness: the subject undergoes a reduction of their body-space and of their possibility space within an unfamiliar environment framed by an uncertain horizon of experience. Equally important, seeing the effect that disorientation has on self-consciousness allows us to see the effect that typical integrated self-location has on self-consciousness: integrated self-location enriches self-consciousness by contributing to the horizon of experience, providing the subject with a set of possibilities for both action and perception. Importantly, disorientation is a degradation in self-location (conceived agentively), and as such, a degradation in self-consciousness (conceived agentively10), but outside of maybe extreme cases (e.g. prolonged episodes of nearly total darkness leading to de-realisation), disorientation does not result in an obliteration of self-consciousness. Borrowing the wording from the discussion of orientation in the work of Erwin Straus (Straus was discussing orientation in dancing), disorientation transforms “the structures of immediate experience” in which self-consciousness is embedded11 (p. 30, Straus 1966).
IV– THE PHENOMENOLOGY OF DISORIENTATION After reviewing existing discussions on the ties between self-location and self-consciousness, I drew a distinction early on between minimal self-location (involving only egocentric frames of reference) and integrated self-location (involving egocentric and allocentric frames of reference). This distinction gave me the necessary framework to explore the phenomenology of disorientation, starting with experiences of being turned around, then studying illusions of orientation along the other two egocentric axes and finally experiences of disorientation that did not necessarily involve illusions along the egocentric axes. I proceeded to show that illusions of orientation can happen along the three egocentric axes, supporting the idea that all three egocentric axes are necessary for integrated self-location. I then argued that integrated self-location is transmodal, having effects on both perception and perceived possibilities for action. In section three, a phenomenological investigation of the feelings associated with disorientation (in particular, BSR and unfamiliarity) led 10 For an early account of agentive self-consciousness see Husserl’s notion of “Ich kann” (§38 of Husserl 1990 and §28
of Husserl 1970). According to Husserl, the subject experiences the world through her lived body as a series of “I can’s”, that is, as a series of possibilities for action.
11 Here, when we think of self-consciousness, we have to remind ourselves that, as Bermudez points out, grasping one’s
possibilities for action within an environment plays a fundamental role in self-location and therefore, in the “self-conscious subject’s grasp of how he himself is a part of the physical world” (p.223, Bermudez 1998).
— 37 —
me to the conclusion that during disorientation the subject undergoes a reduction of its body-space and of its possibility space within an unfamiliar environment framed by an uncertain horizon of experience. The reduction of body-space and possibility space during disorientation suggests that there might be empirically testable links between disorientation and the sense of agency, and between disorientation and peripersonal space. Another potential avenue of research is to study the connection between the different neural mechanisms of spatial cognition and the destabilization of integrated self-location. For instance, in vivo recordings demonstrate that without vision the rodent head direction system (a system of cells within the hippocampal formation that discharge as a function of the directional heading of an animal) becomes unstable very quickly (Cheung et al 2012). It would be interesting to explore if and how the instability of the head direction system relates to the instability of the horizon of experience characteristic of episodes of disorientation. In this chapter, I have focused on what one could deem paradigmatic disorientation experiences, involving feelings of anxiety, confusion, vulnerability and unfamiliarity. However, not all experiences of disorientation follow the same pattern. In our corpus, a small subset of the subjects report experiencing positive feelings when disoriented (“the prospect of being totally lost is sometimes good fun,” “[I felt] joy…it was very cool and [I had] never felt something like that”), sometimes even experienced in conjunction with negative feelings such as fear or confusion. A research question left to explore is how the phenomenology of negative and positive aspects of disorientation experiences differ and in what aspects they are similar. Accordingly, the next chapter will explore under which condition disorientation brings about not a strange and threatening world, but a novel world full of freedom and possibility.
— 38 —
3 . L E A R N I N G H O W T O G E T L O S T : E V E N K I
N AV I G A T I O N A N D ‘ M A N A K A N ’
The question is how to get lost. Never to get lost is not to live, not to know how to get lost brings you to destruction, and somewhere in the terra incognita in between lies a life of discovery.
-p.1, Solnit 2005
I – INTRODUCTION In Geographies of Disorientation, Marcella Schmidt di Friedberg distinguishes between the negative and positive aspects of spatial disorientation — what she calls, respectively, the “uncanny feeling” and the “delicious flavour” of getting lost (Schmidt di Friedberg 2017). That disorientation should be an uncanny feeling comes to most as no surprise: when lost, most people feel anxious, helpless and confused, and the surrounding environment turns unfamiliar and threatening. However, the positive aspects of disorientation are less commonly experienced and more rarely the subject of research. In an effort to bring to the fore some of the positive aspects of disorientation, Schmidt di Friedberg establishes important links between the fields of geography, design, art and anthropology. She points first to the work of French-Swiss designer Ruedi Baur, who holds courses and seminars on the theme of Disorientation and who advocates for getting lost as a practice that establishes a new relationship to space:
Wandering around at random, almost purposely losing track, travelling around, trying to forget where you come from, where you are and where you are heading. […] Moving somewhere else not from obligation but for pleasure […] Doing things with no particular purpose, choosing to venture into unknown lands or cultures […] add one complexity to another, blur distinctions, delve into chaos not to bring in any order but to enjoy getting beyond simplistic models.
-Baur in Schmidt di Friedberg 2017, p. 9
This potential of disorientation to take us “beyond simplistic models” resonates with the ‘dérives’ of the Situationist movement — unplanned journeys in which “one or more persons during a certain period drop their relations, their work and leisure activities, and all their other usual motives for movement and action, and let themselves be drawn by the attractions of the terrain and the encounters they find there” (Debord 1958). According to the anthropologist Franco La Cecla, in a dérive, getting lost is “aimed at restoring value to the undecidable and radically anarchical aspect of spatial experience” (La Cecla in Schmidt di Friedberg 2017, p. 9). Thus, disorientation, according to
— 39 —
Schmidt di Friedberg, holds a potential for spontaneity and freedom, for embracing uncertainty, and for establishing a new kind of relationship with space. The central aim of this chapter is to build an account of the positive aspects of spatial disorientation. The development of this account will build on the analysis of the phenomenology of disorientation presented in the previous chapter and on ethnographic research among the Evenki natives of central Siberia. Evenki reindeer herders and hunters have unique navigation methods that result in a very special relationship to their environment. A central aspect of this relationship is the feeling of manakan (‘making your own way’ in Evenki language), a feeling of autonomy and independence. A study of Evenki navigational style and its relationship to manakan will serve to elucidate the workings behind the emergence of the positive aspects of spatial disorientation. In chapter 2, I established a distinction between minimal self-location and integrated self-location (i.e., in the integration of egocentric and allocentric frames of reference through the online system of spatial representation). This distinction helped me characterise the standard case of disorientation as a destabilisation in integrated self-location resulting in a destabilisation of the subject’s horizon of experience. Spatially, our immediate environment is often embedded within a larger configuration of out-of-sight landmarks. When a Londoner finds herself in Oxford Circus, she has a sense of Regent’s Park, Piccadilly, Marble Arch and Tottenham Court Road hovering somewhere beyond the horizon even if she cannot see them. When disoriented, it is precisely this large-scale spatial frame of reference (i.e. her horizon of experience) that becomes destabilised. The destabilisation of the subject’s horizon of experience leads naturally to the negative aspects of the experience of disorientation (e.g. anxiety), but less so to the positive aspects of disorientation that we just mentioned. There are good reasons for us to follow the idea that in disorientation there is a destabilisation in the integration process of allocentric and egocentric frames of reference and, phenomenally, a destabilisation of the subject’s horizon of experience, but this raises the question of how such a destabilisation can possibly lead to a sense of freedom and spontaneity. Using the terminology introduced in chapter 2, we could rephrase the central research question of this chapter as follows: how can the destabilisation in the integration process of allocentric and egocentric frames of reference ever hold a potential for experiencing a sense of freedom and for establishing a new kind of relationship with space? In the following section, I will introduce the case of Evenki reindeer herders and hunters and discuss their navigational methods, using both our own ethnographic work and previously existing research. Section 3.3 will introduce the mind map vs practical mastery theoretical debate in anthropology and section 3.4 will employ the distinctions arising from this debate to analyse Evenki navigation. Section 3.5 will analyse the central features of the experience of manakan in Evenki culture. Section 3.6 will study experiences of disorientation among Evenki hunters and reindeer herders. Section 3.7 will link manakan to the positive aspects of spatial disorientation. Section 3.8 will provide a conclusion and potential avenues for future research.
II – INTRODUCING EVENKI NAVIGATION The Evenki, a Tungusic people spread over northern Asia, are renowned for their hunting and navigational skills (Vasilevic 1963, Lavrillier 2006). In 2019, I went on a winter expedition and on a summer expedition to the Evenkia region of Krasnoyarskii Krai (Russia) in arctic and subarctic Siberia. During that time, my expedition teammates and I interviewed reindeer herders (in Surinda during our winter fieldwork) and ex-reindeer herders and semi-nomadic hunters (in Chirinda during our summer fieldwork). Our research included interviews and, to a lesser degree, participant
— 40 —
observation. The interviews were semi-structured, in Russian, and no questionnaires were used. The interviews were generally prompted with a question that then led to a discussion. The present chapter builds on our collected ethnographic material as well as on existing ethnographic research to investigate the Evenki methods of navigation and how they relate to the feeling of manakan. Regarding the apparent divide between the nomadic and sedentary Evenki in our study (e.g. reindeer herders vs hunters), I follow Vladimir Davydov’s analysis, in which he argues that the Evenki way of life should be understood through the prism of everyday practices and their related patterns of mobility, rather than through a populational divide between nomadic and sedentary Evenki (Davydov 2013). After all, both herders and apparently settlement-based Evenki engage in nomadic practices (Dumont 2018). Reindeer herding, although still present, is no longer the majority occupation among the Evenki in the regions that we visited. Still, the lives and practices of the majority of the Evenki we met could hardly be characterised as sedentary. Evenki hunters leave their settlements for hunting seasons that span several months. And in lake Chirinda, we stayed with an old couple of ex-reindeer herders that lived in a chum (the traditional itinerant dwelling of reindeer herders, which is the Evenki equivalent of a yurt) at the other side of the lake from the settlement, moving their dwelling with the seasons. The navigational practices of the Evenki developed over a time in which reindeer herding was the main mode of life of the Evenki people. Nevertheless, the navigational practices that I will be discussing are present in both nomadic reindeer herders and semi-nomadic hunters. Nowadays, the Evenki have reasonable access to modern technology, such as snowmobiles and cell phones. Despite this possibility, most Evenki do not carry GPS devices or maps with them when they are navigating the taiga, not even a compass. The Evenki we encountered regarded these cognitive artefacts with scorn, as something of a nuisance or, at best, a game, that had little to do with walking in the Taiga. Anecdotally, we witnessed a reindeer herder winning a surprise prize in a local festivity and being crestfallen to discover it was a GPS device and not the portable radio that he had hoped for. A quote by the reindeer herder Oleg Iakovlev (as recorded by Lavrillier and Gabyshev) reflects well the Evenki attitude towards external representations of the environment: “Everything is written in the natural environment, you just need to observe all around you” (p. 15 Lavrillier and Gabyshev 2017). At the heart of the Evenki way of navigation is what many ethnographers refer to as skilful walking. After long-term ethnographic research, Brandišauskas highlights the importance of skilful walking in Evenki practices: “dwelling on the taiga is based on skilful attunement to a place…while moving through it” (p. 23 Brandišauskas 2012). This skilful walking is embedded within a complex network of paths and takes place in the context of vast hydrological knowledge and a rich toponymical system (i.e. a system of names of places). The paths (or tracks, as they are often invisible to the untrained eye) are created and maintained with the movement through the taiga of herders and hunters, and although the network was much broader during pre-soviet times, it is still an important element in Evenki navigation (see Campbell 2001 for a detailed analysis). The Evenki have an extremely detailed knowledge of the area where their usual patterns of movement take place, and a knowledge of the river systems of a much larger area:
During my missions in the thick taiga forest between 1925 and 1961, the Evenks, whether they learned to write or not, always drew me maps showing only the rivers and all their tributaries, measuring distances in days of travel. Most of them kept in memory the complete scheme of a river, but also all the meeting points between the great rivers. Of course, none of them orientated their drawing like our geographical maps, according to the absolute cardinal points; on the other hand, they were never mistaken about the direction of the rivers' currents. Before drawing the map, they would turn the sheet of white paper in
— 41 —
their hands for a long time, and then, once they had found the right position that would contain the space to be represented, they would begin by drawing the most important river, running the pencil from them in the direction of the current, and gradually tracing all the tributaries of this river, and also those of the nearby river to express the presence of a pass.
-p. 315 Vasilevič 1963
Alexandra Lavrillier remarks how during map-sketching exercises the Evenki were never interested in cardinal points or in landmarks such as mountains, but in rivers (Lavrillier 2006). She later noticed that adults and children alike recited large rivers and their tributaries as a pastime, the way one might recite multiplication tables. This detailed hydrological knowledge supports the capacity of the Evenki to explore territories that go well beyond their habitual lands of herding or hunting. Rivers also serve to delineate different regions and thus to support everyday navigational activities. When they are talking or sketching itineraries, the Evenki always take as a point of departure the location of their tent in relation to its closest river and describe walking movement in reference to the direction of the current. Knowing the contour of nearby rivers means that when hunting, the rivers can always be used as a return path and as borders of different areas. Evenki also have a detailed knowledge of landscape names (i.e. toponomy), both of particular sites and of types of landscape. This typology of landscape is extremely useful, because place names are very specific (e.g. ‘Ily’ means a piece of rock sticking out of a mountain) and they carry information about the possible uses of the landscape. This knowledge system is dynamic and relational: Evenki can use this typology to infer the types of landscape that will surround a particular landscape type (see Lavrillier and Gabyshev 2018 for a detailed analysis of Evenki landscape typology and a glossary of place names). In the words of researcher and reindeer herder Semën Gabyshev, “thanks to the Evenki knowledge system of the landscape, I can guess the locations of rocks, rivers, passes, and so on, even if I do not know the specific place” (p. 25, Lavrillier and Gabyshev 2018).
III – MENTAL MAPS VS PRACTICAL MASTERY It is fruitful to frame the discussion of Evenki navigation within the context of a major current debate in anthropology, that between proponents of the “mental map” theory of human navigation and the “practical mastery” theory of human navigation12. The interesting distinctions of this debate will help advance our understanding of Evenki navigation and, in turn, a more detailed understanding of Evenki navigation can help advance the debate. The “mental map” theory holds that wayfinding “is carried out in the light of stored spatial information in the form of a ‘mental map’ of the terrain, plus, presumably, some inferential schemes of converting this information into suitable practical decisions and actions” (p. 272, Gell 1985). In contrast, “practical mastery” proponents defend that most forms of wayfinding don’t rely on abstract representations of spatial relations, and that we should understand wayfinding “as a skilled performance in which the traveller, whose powers of perception and action have been fine-tuned through previous experience, ‘feels his way’ towards his goal, continually adjusting his movements in response to an ongoing perceptual monitoring of his surroundings.” (p. 220, Ingold 2000). The contrast between the views of Edwin Hutchins and Tim Ingold on the topic of navigation provides a useful illustration of the two positions:
When the navigator is satisfied that he has arrived at a coherent set of correspondences, he might look to the chart and say ‘Ah, yes; I am here, off this point of land.’ And it is in this
12 In this chapter I use navigation and wayfinding interchangeably, but other authors make a difference between these
two terms by conceiving navigation as aided (i.e. by cognitive artefacts) and wayfinding as unaided.
— 42 —
sense that most of us feel we know where we are. We feel that we have achieved reconciliation between the features we see in our world and a representation of that world.
-p. 13 Hutchins 1995
‘Where am I?’ is not ordinarily answered in terms of a location in space, determined by the intersection of an independent set of coordinates. Hutchins to the contrary, it is not in this sense that most of us feel we know where we are. Indeed, I may know precisely where I am and yet have no idea of my geographic location. For it is not by assigning the position where I currently stand to certain spatial coordinates that an answer to the ‘where’ question is arrived at, but rather by situating that position within the matrix of movement constitutive of a region.
-p. 237 Ingold 2000 Central to the debate is the concept of non-indexicality, which “corresponds with the notion of an allocentric as opposed to egocentric frame of reference in psychology” (Istomin and Dwyer 2009; see also chapter 2 for a discussion of allocentric versus egocentric frames of reference). Using this notion of non-indexicality, Gell defines maps as “any system of spatial knowledge and/or beliefs which takes the form of non-token-indexical statements about the spatial locations of places and objects” (Gell 1985, 278)13. Thus, according to mental map advocates, wayfinding depends on placing the subject in a non-indexical system of spatial knowledge. That is, on integrating egocentric and allocentric frames of reference and, as a result, reconciling “the features we see in our world and a representation of that world” (p. 13 Hutchins 1995). Against this view, practical mastery advocates defend that during wayfinding there is no computation of one’s location with respect to a non-indexical representation. To support their position, they refer to an array of anthropological research coming from work with different indigenous groups. A member of one of these groups might be unable to specify their location in space within a non-indexical representation of a region (or a system of coordinates) and yet they are still adept at wayfinding in that region. Additionally, practical mastery theory builds on the insights of the Gibsonian ecological approach to psychology, in particular on Gibson’s theory of ‘reversible occlusion’, the central elements of which are vistas and transitions between vistas (p. 198 Gibson 1979). The set of visible surfaces along a path constitutes a vista. Importantly, a vista is not a picture seen from a static position, but a dynamic set of surfaces that is created by moving along a path. Imagine walking down a busy shopping street and passing different stores on your right and the road to your left. The individual stores might fall behind you as you walk, but it is the joint flux of passing surfaces both vertical (e.g. stores) and horizontal (e.g. the road) along this path that constitutes the vista. Now, if you turn onto a different street, there is a transition, which opens another vista (i.e. that of the surfaces along that new street). Following Gibson’s line of research, the psychologist Harry Heft argues that wayfinding depends on moving through routes that recreate the flow of perspective particular to the path leading to the destination (Heft 1996). Tim Ingold refers to this psychological research in order to argue that wayfinding depends not on reflexive planning and location updating based on mental maps, but on an engaged series of transitions between vistas as one walks — in a phrase that summarises well the practical mastery position, “we know as we go, not before we go” (p. 239, Ingold 2000). Istomin and Dwyer review the map vs mastery debate and argue that it corresponds to the survey and route knowledge distinction (Shemyakin 1962; Appelyard 1970; Siegel and White 1975). Navigation based on mental maps corresponds to survey knowledge (non-indexical knowledge of 13 This is arguably not the best definition of a map (and not one I would necessarily endorse), but for our purposes, it
serves as a stipulatory definition to clarify the “mental map” position in the debate.
— 43 —
spatial locations) and practical mastery correspond to route knowledge (knowledge of a network of routes). While the survey knowledge and route knowledge distinction is useful to reframe the debate at hand, one should be wary of it — they might be best conceptualised not as types of spatial knowledge but as styles of describing a space, and even then subjects switch between styles when describing a given space (for a nuanced discussion on the structure of spatial knowledge, see Tversky 2000). Still, this connection is important because it points out that transitions between vistas alone cannot explain many of the features of human navigation that are accounted for in the mental map model of spatial cognition. Most notably, mental maps enable generalisations and inferences that go beyond the spatial information gained through direct experience (Allen 1999). Imagine that A, B and C form a triangle within a street network. Imagine now that an individual had learned the route from A to B and from B to C, but not from A to C. If the individual were to use only route knowledge to go from A to C, they would have to go from A to B and then from B to C. In contrast, using survey knowledge (e.g. picturing the location of A, B and C forming a triangle), they would infer that there is a more direct route from A to C and follow the path closest to the direction to C from A. The capacity for these kind of inferences (e.g. taking a shortcut, taking a detour…) is found across cultures, and this presents a substantial problem for the ‘practical mastery’ position. Istomin and Dwyer contend that rather than conflicting, the two theories in the debate are complementary. Human beings both use mental maps and route knowledge during navigation. What varies both across cultures and across individuals is the degree to which one of the two types of knowledge (survey vs route) dominates the other. They then re-evaluate the debate in the light of ethnolinguistic studies. Linguistic expressions of spatial relations can be given in relative (e.g. the bakery is to the left of the café) or absolute frames of reference (e.g. Times Square is south of Central Park). The prevalence of the use of the different linguistic frames of reference varies across cultures, with some languages only allowing a single frame of reference. For example, the Australian language Guugu Yimithirr only permits the absolute frame of reference (Levinson 2003). Intercultural studies at the intersection of ethnolinguistics and spatial cognition consistently found that differences in linguistic frames of reference tended to result in differences in navigational styles, with “absolute people” more likely to rely on survey knowledge and “relative” people on route knowledge (Istomin and Dwyer 2009). Interestingly, when asked to sketch a map, “absolute people” will orient the map according to an absolute coordinate system (e.g. geographic bearing) and “relative people” will sketch non-oriented maps. Of course, this absolute-relative people distinction is only part of the story. Istomin and Dwyer did studies with reindeer herders belonging to two different groups (the Komi and the Nenets) and found that even though they are both “relative” people (linguistically) and have roughly similar ways of life, their mental maps and the degree to which they rely on them for navigation varies.
IV – AN ANALYSIS OF EVENKI NAVIGATION Following the distinction introduced in the previous section, it seems clear that the Evenki would fall on the category of “relative” people. This is not only clear linguistically, but also in terms of them relying mainly on route knowledge (e.g. the paths mentioned in the previous section) and having a strong tendency to sketch non-oriented maps. The Evenki have a vast hydrological knowledge and they can build on that knowledge to sketch maps at an ethnographer’s request, but these maps are oriented in reference to rivers, not to absolute coordinates (Vasilevic 1963). For the Evenki, “cardinal points don’t constitute a real point of reference” (p. 23 Lavriller 2009).
— 44 —
Our interviews confirm that Evenki seem to rely largely on route knowledge when they walk in the taiga. Their constant refrain about how to orient is that the Evenki just know the land, and that they find their way by walking. And when talking about walking in the taiga, they often made reference to the presence of tracks that they could follow to different places. When we asked an Evenki hunter what he would do if lost, he looked at us confused and said “well, I would just find my way”. When we inquired about the use of maps in particular, another Evenki hunter, who had been drafted and undergone military training, told us that he had learned how to use a map and compass there. He had found the whole exercise quite fun and had become fairly skilled at locating himself in a map using a compass and landmarks. When we asked him if that was similar to wayfinding in his hunting grounds, he looked at us a bit perplexed and explained that maps and compasses were fun, but that they had nothing to do with walking in the taiga, because to walk in the taiga you need to know the land. An interview with a Slavic hunter who hunted in the same region provides an illustrating contrast. We simply asked him how he oriented while hunting and he gave the following detailed description, which matches very well the use of survey knowledge:
Well, how could I say it? I have this map in my head, you see. Of the whole region where I hunt and a bit further on. A very detailed map, with all of the mountains and with how the crests and valleys are oriented. And then, I keep track. I look at the mountains and I know which ones they are, and I know where to go. And I always know where north is, I always keep track. Always. Even in the snowstorms. I know in which direction the wind is coming, I feel it at a certain angle, and I keep that angle in order to advance in a straight line.
While the Slavic hunter relied clearly on survey knowledge (mind maps), the Evenki hunters operating in the same area seemed to rely mostly on route knowledge (practical mastery). Importantly, this does not mean that the Evenki were not able to use maps or to map some of their knowledge, just that the ability to locate themselves within an abstract non-indexical representation of the space around them is not central to their wayfinding processes. Moreover, the ‘skilful walking’ (skilful attunement to a place while moving through it) so central to the Evenki ethos, seems to be very much in line with the idea of practical mastery. Another aspect of Evenki walking that points in the practical mastery direction is that they prefer to have no predetermined destination and to instead open themselves to the possibilities presented by the environment (Safonova and Santha 2013), which resonates with the previously mentioned idea that they know as they go, not before they go. This still raises the question of the role of allocentric and egocentric frames of reference during wayfinding for the Evenki. Knowledge of paths and skilful walking don’t seem to require allocentric frames of reference. This lack of allocentric frames of reference while being immersed in the taiga seems to be patent in their phenomenology of space as well. Contrast a Londoner walking through Oxford Circus who has a sense of Regent’s Park, Piccadilly, Marble Arch and Tottenham Court Road hovering somewhere beyond the horizon with the following description of the Evenki experience of skilful walking: “territory is perceived as a field of open possibilities, of potential directions always in relation to the person’s location at any given moment” (p. 66, Shirokogoroff 1935). This is very close to Ingold’s idea of the matrix of movement that constitute a region. However, there is a difference in emphasis. Ingold emphasises the histories and memories of previous movement: “For those who know a country, in short, the answers to such basic questions as ‘Where am I?’ and ‘Which way should I go?’ are found in narratives of past movement” (p. 237, Ingold 2000). This is of course an important part of Evenki experience, with paths having histories and places having associated memories. Nevertheless, I would like to bring the emphasis from
— 45 —
remembered movement to potential movement. Rather than remembering previous movement, the way the Evenki feel the space is as a matrix of potential movement. The above applies neatly to two of the central elements of Evenki navigation introduced in section 2. Namely, the Evenki system of paths and their skilful walking. There are two other elements, however, that at first seem at odds with navigation based on route knowledge: hydrological knowledge and the Evenki toponomy of landscape. The Evenki have a detailed knowledge of river systems, going from large rivers to small tributaries over very large areas. Even though this knowledge might be better described as narrative-based, or at best, diagram-based (something akin to a genealogy tree), ethnographers have found that the Evenki are able to draw non-oriented maps in which their itineraries are sketched in reference to river systems (see figure 1 below; and Lavrillier 2006 for further examples). During wayfinding, the Evenki rely on rivers as return paths and as borders. Qua return paths, rivers should be no different than other paths in that they need not involve survey knowledge. What is interesting is that rivers work for the Evenki as “lines that delimit both hunting grounds and herding grounds and [on a larger scale] the region of nomadic movement of a clan” (p. 19, Lavrillier 2006). Lavrillier gives as an example the itinerary of a hunter who is pursuing a prey in an area between the Tungurca and Umuksa rivers (see figure 1 below). The two rivers run roughly in parallel 10 km apart and the area is approximately 30 km across. After finding the trail of the animal, the hunter first goes down to the Umuksa river, then meanders diagonally across until he reaches the Tungurca river, then in a descending diagonal until he ends on a tributary of the Umuksa, crisscrossing as a result the actual animal itinerary which goes horizontally in between the two rivers. He follows the tributary until finally finding the prey, shooting it and hiding the carcass under a blanket of snow for later transport. Then he follows again the tributary all the way to the river Umuksa, which he uses as a return path to his tent (chum), set up on the bank of a further tributary of the Umuksa. And if the hunter then wants to go back to the carcass, he just needs to retrace his footsteps from the tributary.
Figure 2. Adapted from Lavrillier 2006
— 46 —
This hydrological knowledge should not be simply equated with a system of landmarks in an allocentric frame of reference. The rivers (as borders) serve to delimit the areas of activity of the Evenki during wayfinding. As such, they serve as a sort of safety net during skilful walking. The Evenki can immerse themselves in the environment egocentrically and they will always be able to find their way back to their tent. As a result, it seems that there is no set of distant spatial locations that hover beyond the horizon of experience of the Evenki. And as with many wayfinding systems, there is a multiplicity of methods that the Evenki can use to go back to their tent after being immersed in the environment egocentrically, which contributes to what I would deem a redundancy-based navigational robustness (e.g. Cherubini and Chaumette 2010). Evenki navigators can simply backtrack, retracing their own itinerary backwards; they can use path integration; they can rely on the large system of paths and follow the route that takes them to their tent; they can follow rivers; and they can use their typology of landscape for wayfinding inferences. A further interesting aspect of rivers from the point of view of cognitive ethnography is that rivers do not have to be visually perceived directly to be exploited for navigation. The surrounding landscape (e.g. slope, vegetation) indicates the location of a not-yet-visible river, and the sound of the water can provide an auditory cue of the river’s location even in densely forested areas where visibility is reduced. As for the typology of landscape, we saw in section 3.2 that the Evenki have names for and recognise many types of landscape. They can use this knowledge to know the possibilities (for wayfinding and otherwise) of each type of landscape (e.g. whether it is easy to climb) and they can also guess fairly accurately the types of landscape that will surround a given landscape type, which as I just mentioned, can serve them to find rivers, or to find the section of landscape in which their tent is based. However, all of this knowledge and methods of inference do not seem to provide anything resembling survey knowledge. Together, skilful walking, path systems, hydrological knowledge and typological knowledge provide a way for the Evenki to navigate without relying on survey knowledge. When hunting, this navigational system allows them to “forget all concerns… in order to immerse themselves in the situation” (p. 103, Safonova and Sashta 2013). This is incredibly important for the Evenki, because, as will become clear in the next section, it is essential in order to cultivate the feeling of being ‘manakan’.
V – MANAKAN In Evenki culture, manakan denotes a feeling of autonomy. During moments of manakan, self-reflexivity falls away and the Evenki experience a unique sense of freedom. The term manakan is used both as a feeling and as a way of being. Sometimes, Evenki refer to feeling manakan, and other times, they refer to someone being manakan or displaying manakan. For example, the more sedentary individuals living in settlements (e.g. public servants) would tell us that such and such individual was manakan, usually referring to individuals that spent a long time immersed in the taiga. The etymology of manakan (to make one’s own way) points to its connection to wayfinding. Based on their extensive fieldwork, Safonova and Sashta argue that this experience of autonomy is central to the development of the Evenki ethos and “comes from the ability to walk and travel alone… Landscape also plays an integral part in this ethos, because the areas through which one can walk alone determine the extent and intensity of manakan experiences.” (p. 159, Safonova and Sashta 2013). Only thanks to the complex navigational system of the Evenki are they able to immerse themselves in the taiga and feel this sense of autonomy. In order to feel ‘manakan’, the Evenki need to liberate themselves from plans and expectations. That Evenki shy away from planning is a point of agreement of many ethnographic accounts:
— 47 —
“Hunters do not like to predict walking time and [an Evenki hunter] used to tell me ‘we should not try to guess’ (R. ne budem zagadyvat’). This also relates with hunting beliefs that beings can react to what is planned in advance” (p. 199, Brandišauskas 2009). This lack of plans and expectations facilitates manakan because it means that the Evenki are open to “any options and possibilities, however various and unpredictable, that might arise.” (p. 115, Safonova and Sashta 2013). Manakan constitutes not only a distinct way of experiencing oneself (as autonomous and free) but also a distinct way of experiencing the environment. During the experience of manakan “one perceives places in a unique way” (p. 142 Safonova and Sashta 2013). Here we draw on the previously made point that the Evenki, thanks to skilful walking, perceive the environment as a field of open possibilities and potential directions. Thus, manakan should not be considered as freedom in an atomistic sense, but in a relational sense. Manakan is not a way for the Evenki to impose their will on the environment. On the contrary, manakan opens the possibilities of the landscape to the Evenki in a way that transcends the human abstractions of plans and expectations. It is this feeling of possibility that underlies the experience of autonomy. As a result of skilful walking, the Evenki perceive the landscape as alive, as capable of responding to their actions (Sirina 2008). It is this playful attunement to a living environment that engenders the feeling of manakan. Ironically, we can further our understanding of the feeling of manakan by studying its absence. The Evenki never conceive space as closed and determined; they need to conserve a “freedom of spatial evolution” (p. 3, Lavrillier 2006). As a result, the Evenki in the settlement “long for the freedom, self-autonomous organisational and spatial order of the big forest” (p. 23, Mustonen and Lehtinen, 2020). In contrast to the taiga, the towns are seen to be ordered according to Russian norms, both politically and spatially. The result is a draining away of manakan during sedentary practices. The alienation of the Evenki under the Russian sedentary way of life is perhaps most clear in their attitude towards the spatial routines of the settlement:
Having a set purpose … spoils the pleasure of being on the road and prevents one from being fully involved in the journey. Whenever possible, Evenki avoid travelling in such conditions ... When they finally start their journey home, the first chance encounter will stop them in their tracks and they will accompany whomever they met back to the village. If there is no chance of escaping from a trip with a predetermined aim they will be sullen and uncommunicative, as if the existence of this concrete purpose prevents them from feeling free and accepting the pleasure of the trip.
-p. 143, Safonova and Sántha, 2013
The delight of the Evenki at chance encounters is something that we experienced often when visiting settlements. On a daily basis, we would meet someone, and they would insist that we accompanied them. We would walk them all the way to their house, and upon arrival, they would insist on accompanying us to our house. If more people were encountered on the way, this would be a further cause for joy, as it would result in new unexpected switches in the itinerary. Soon enough, we started to partake in the pleasure of finding ourselves in groups that gained and lost members as we wandered through the settlement. As Safonova and Sántha explain, all of these chance encounters “liberate the traveller from the prescriptiveness of the initial purpose of a journey and allow him or her spontaneously to change the route, and combine different tasks and possible outcomes. This sense of having multiple possibilities fills the situation with joy.” (p. 143, Safonova and Sántha, 2013). Prescriptiveness is a key word here. All of the Evenki cultural practices of movement aim to rid the experience of the environment of any prescriptiveness. Consequently, the Evenki have a renowned passion for exploring new territories, often for months and without a clear instrumental purpose
— 48 —
(Lavrillier 2006). And in Transbaikalia, they celebrate an annual ritual in which they wander between secret places without any prescribed movement, often walking in circles and making loops (Safonova and Sántha, 2013). In our interviews, reindeer herders told us that they sometimes left the decision-making during walking to the reindeer, which ensures moving in non-prescribed ways. Hunters reported doing likewise when chasing a prey with their dogs. A hunter in Chirinda described it in the following way: “The dog runs here and there, and as we go, everything is surprising and exciting”. Manakan requires the ability to walk alone, which is deeply connected to the Evenki system of navigation. At the most basic level of analysis, an Evenki experiencing manakan still needs to be able to make it back to their tent at the end of the hunt, and this is made possible by the network of paths, hydrological knowledge and the use toponymical-based inference. What is fairly unique about the Evenki method of navigation is that it allows for wayfinding that is not prescriptive, which is essential to the attainment of manakan. Here again, let us contrast the Londoner in Oxford Circus with the Evenki hunter in the taiga. While the horizon of experience of the Londoner is framed by out-of-sight landmarks, the horizon of experience of the Evenki is a matrix of open possibilities. As we have seen, Edmund Husserl clarifies that the horizon of experience “is not a nothingness, but an emptiness to be filled out; it is a determinable indeterminacy” (p.41, Husserl 2012). Of course, this determinable indeterminacy can be determined to different degrees. The abstract spatial configuration of the Londoner’s horizon of experience is (comparatively) overdetermined and it generates a set of expectations about the environment. In comparison, the Evenki is surrounded by uncertainty. The immediacy of the Evenki experience of the environment contributes to the unpredictability and risk necessary for feeling manakan and to the perception of the landscape as alive and capable of responding to human action. This unprescribed experience of the environment is only made possible because for the Evenki there is no ongoing integration of egocentric and allocentric frames of reference during wayfinding — no reconciliation between the features we see in our environment and a spatial representation of that environment, following Edwin Hutchins’s terminology. The integration of egocentric and allocentric frames of reference can assist navigation for people relying chiefly on survey knowledge, but it also overdetermines the navigator’s horizon of experience through expectation and anticipation. It is precisely this integration that the Evenki go to great lengths to avoid. Skilful walking makes possible for them to navigate without the continuous integration of egocentric and allocentric frames of reference, i.e. without constantly updating their position in some mental map of the environment. In turn, the skilful attunement to the environment in Evenki walking is only sustainable thanks to a complex navigational system (involving path networks, and vast hydrological and toponymical knowledge) that assures that the Evenki are always able to find their way back to their point of origin. The result is that the Evenki experience themselves as free individuals moving through an environment that is alive and rife with possibility.
VI – EVENKI NEVER GET LOST If the subject of inquiry is disorientation, it might seem odd to study a group of indigenous people who are famous for never getting lost. And indeed, when we asked Evenki hunters and reindeer herders if they had ever gotten lost, the reactions would range from laughter to affront, followed by the matter-of-factly assertion that Evenki never get lost, as if disorientation were a particularly Slavic affliction. Safonova and Sántha had a similar experience in their ethnographic work: “We asked Irgichi if the Evenki ever got lost in the taiga. He laughed and told us that Evenki could only become lost if they were injured and lost their minds” (p. 141, Safonova and Sántha, 2013). The
— 49 —
two ethnographers concluded that as the Evenki never know in advance where they will go, they never lose their way. Safonova and Sántha to the contrary, Evenki can get lost. However, this is a most extraordinary event. I had to go back 100 years until I finally found an ethnographic account of an Evenki getting lost, in the works of the early 20th century anthropologist Maria Czaplicka (Czaplicka 1916). Czaplicka had gone on a fieldwork trip to the same region of Evenkia that we were in, and she travelled from chum to chum all the way to Chirinda, following different guides. In one of these trips, Czaplicka and her guide suddenly found themselves in a snowstorm, and after a while, the Evenki guide lost his way. After declaring that he was lost, the Evenki was nervous and extremely upset. After calming down somewhat, he decided to leave Czaplicka with the reindeer and go exploring in order to reorient. Eventually, he found a path that led to a river. He backtracked to find Czaplicka and after reaching the river they found the chums of another group of reindeer herders. It was only towards the end of our second fieldwork season that an Evenki hunter confided in us another account of having gotten lost. One can expect that Evenki are reticent to share such accounts out of shame or pride, but I believe that the main reason that there are so few existing accounts in the ethnographic literature is truly that Evenki almost never get lost and that most of them have never experienced getting lost. As it happens, the hunter that told us about this incident said that nothing similar had ever occurred to him in his life. He had been hunting for several days, getting further and further from his usual hunting grounds as a result. He had been following sable for a prolonged period when the hunt led him from dense woods into an open valley. He suddenly realised that the place was not one he recognised. Here, he describes a most interesting phenomenology: “I felt a pang of panic, and it was as if a great fog was descending upon me from the hills” (he made a gesture of his open hands closing against each other to illustrate the feeling). In his panic, he moved restlessly up and down, until he managed to force himself to calm down. He made a fire and sat by it while thinking what to do. Eventually, he went back into the woods and wandered until he found a river, which he followed back to his camp. We have then two accounts of Evenki getting lost. The two are uncommon situations. One of them involves being caught in a snowstorm with zero visibility. The other one involves unexpectedly finding oneself in a totally unfamiliar landscape. In both cases, the Evenki are deeply affected emotionally (extremely upset and panicked, respectively). And in both cases, the Evenki react by setting out to find rivers. In the first case, by wandering until finding a path within the path network. In the second case, by inferring the location of a river from types of landscape (we asked the hunter how he had found a river and he told us that if you look around, you know where there is a river). The account that we collected points to similarities with western accounts of disorientation. First, the feeling of the environment closing in like fog is strongly reminiscent of the feeling of shrinking (body space reduction) in the analysis of the phenomenology of disorientation in chapter 2. Second, the initial panic and pacing up and down is consistent with the random walking that often ensues when subjects get lost in the wilderness (Hill, 1998). In general, the Evenki wander through regions that they are familiar with during hunting, so, of course, it comes as a shocking surprise to find a completely unrecognisable valley. This shock and the ensuing disorientation prompt the sense of a sudden closing off of possibilities. Similarly, the Evenki in the snowstorm of Czaplicka’s account suddenly finds himself without possibilities for movement, because there is an almost complete lack of visibility and no perceivable paths or landscape types. Just like there are similarities, there is an important difference between the disorientation described in chapter 2 and the disorientation experienced by the Evenki: For the Evenki, disorientation does not seem to follow from a destabilisation in the integration of allocentric and egocentric frames of reference. When subjects rely mostly on survey knowledge, the integration of allocentric and
— 50 —
egocentric frames of reference generates the possibilities for navigation experienced in the subject’s horizon. However, when subjects rely mostly on route knowledge, the possibilities for navigation experienced in the subject’s horizon do not come from the integration of allocentric and egocentric frames of reference. It is the skilful walking of the Evenki that generates a field of open possibilities. When these possibilities close off, their experience of disorientation has similar features to ours, even if this disorientation experience originates from an impairment of different processes (i.e. allocentric-egocentric integration vs skilful walking). Moreover, one can expect that these features are more accentuated for the Evenki, because of how unusual the experience is for them and how much is at stake. After all, the consequences of getting lost in Oxford Circus are not nearly as dire as those of getting lost in the winter taiga.
VII – THE JOY OF DISORIENTATION What the study of Evenki wayfinding can bring to the understanding of disorientation lies not so much in the limited ethnographic accounts of Evenki getting lost, but in the links between the phenomenology of manakan and the phenomenology of the positive episodes of disorientation. Both during skilful walking and during “joyful disorientation” the lack of an integration of egocentric and allocentric frames of reference leads to a sense of freedom and to a radically different way of experiencing the surrounding environment. A comparative study of manakan and disorientation will allow us to understand the central question of this chapter: how can the destabilisation in the integration process of allocentric and egocentric frames of reference ever hold a potential for experiencing a sense of freedom and for establishing a new kind of relationship with space? Let us first revisit some of the features of the positive aspects of disorientation. Ruedi Baur talks about trying to forget our origin, location and destination, about moving with no particular purpose and about getting beyond simplistic models. Guy Debord proposes suspending usual ways of movement and letting oneself be drawn by the attractions of the terrain. And Franco La Cecla defends that getting lost can be aimed at restoring the undecidable aspect of spatial experience. Finally, in her review of the subject, Marcella Schmidt di Friedberg talks about the potential of disorientation for establishing a new kind of relationship with space, so that getting lost may become an experience of freedom. In sum, the positives aspects of disorientation are linked with:
• a lack of origin, location and destination, • having no set purpose, • getting beyond simplistic models, • suspending usual ways of movement, • getting drawn by the attractions of the terrain, • the undecidable aspect of spatial experience, • establishing a new kind of relationship with space, • an experience of freedom.
The idea of forgetting origin, location and destination makes sense when we consider navigation based on survey knowledge, in which moving through the environment is akin to plotting one’s position on a map while trying to go from a point of departure to a goal. Consequently, forgetting origin, location and destination is what happens when one abandons navigation based on survey knowledge. Baur links this with moving with no particular purpose. Compare this with the Evenki notion that having a set purpose spoils the pleasure of being on the road, because it prevents one from being immersed in the journey. Having a particular purpose curtails the possibilities that might
— 51 —
arise from direct engagement with the environment. Moreover, the study of Evenki culture allows us to understand the idea of getting beyond simplistic models better. During orientation, we reconcile the features of the environment with an abstract spatial model. During disorientation, we transcend that model. As a result, there is a potential for establishing a new kind of relationship with space (as Schmidt di Friedberg suggests) and engaging differently with the possibilities of the environment. Abandoning navigation based on survey knowledge is a central aspect of what Debord refers to as suspending usual ways of movement, so it comes as no surprise that the result is that one gets drawn by the attractions of the terrain. There is a deep resonance here with the Evenki hunter who forgets all concerns in order to immerse himself in the situation and be open to any possibilities that might arise. The terrain, both for the Situationists and for the Evenki, is perceived as a field of open possibilities. The lack of an integration of egocentric and allocentric frames of reference, together with the lack of plans and destination, provides for a radically different way of engaging with our surroundings. Debord was careful to point out that dérives should not be strolls led simply by chance, but a process in which the walker is “drifting through a territory to investigate its forces of attraction” (p. 6, Shukaitis, 2014). Just like with skilful walking, this radical switch in modes of relating to the territory can be considered as a skilful attunement to place while moving through it. In this form of engagement, our pattern of attention changes and we become much more aware of our immediate surroundings, now rife with unexpected possibilities. During disorientation, there is a reckoning with what La Cecla calls the undecidable aspect of spatial experience. What lies beyond the horizon is always uncertain, even if it can be determined with different degrees of confidence (i.e. it is a determinable indeterminacy). The integration of allocentric and egocentric frames of reference gives us a sense of certainty in our spatial experience. When we are in Oxford Circus, we have a sense of where Tottenham Court Road and Bond St are. In the Aesthetics of Lostness, the writer Ray Bradbury warns us against the danger of spaces that are too well known and against the lack of enchantment of cities that are “too easily solved”:
There is no mystery, no imaginative lure, no texture in the blank facades, the empty and expressionless faces of banks or other corporate structures. This plus no candy, book, or sweet shops means that when the bank shuts and the IBMers gas on home, the city drops dead. You can skate through those places at 50 miles an hour, for there is nothing to see, nothing to wonder about or linger over. There is no bafflement. The blind buildings do not even hold out a tin cup, asking for dimes. So, because no texture, no attraction, no chance for you to be enticed into even trying to get lost.
-p. 64, Bradbury, 2017
The danger of matching our spatial representations with the features of our surrounding environment is that we can find ourselves surrounded by an environment with no texture and no attraction. In contrast, the Evenki hunter is surrounded by uncertainty, and this contributes to his perception of his surroundings as a living landscape that is at all times changing around him, inviting him to move and reacting to his actions. Likewise, the joyfully disoriented individual dives into the uncertainty of an alive environment. Tim Ingold provides a phenomenological description of what it is to move through an environment in such a way: “The world of our experience is a world suspended in movement, that is continually coming into being as we – through our own movement – contribute to its formation” (p. 242, Ingold 2000). This is what it means to establish a different relationship with space through disorientation. In Schmidt di Friedberg’s words, “getting lost becomes a practice of resistance to the uniformization of landscapes, to abandonment of the daily ritual of getting to know, producing and organizing places” (p. 9, Schmidt di Friedberg 2017). Seen in this way, disorientation holds the potential not
— 52 —
only for anxiety and helplessness, but also for the freedom we experience when we transcend our abstract spatial representations and discover new possibilities in our immediate environment. The generation of survey knowledge and the integration of allocentric and egocentric frames of reference is our main way of getting to know, producing and organising places. This integration of egocentric and allocentric frames of reference makes us experience a space that is determined, organised, controlled. Part and parcel of this spatial organisation is the traditional importance of maps in our culture, and, nowadays, the growing reliance on GPS-assisted navigation. When this spatial organisation falls away (i.e. when allocentric and egocentric frames of reference are not integrated), two things can happen: We might feel the space closing in around us (a feeling of shrinking akin to the one described by the Evenki hunter in an unfamiliar valley) — a reduction of our bodily space and of our possibilities. Alternatively, we might feel a sense of liberation — a rediscovery of our immediate surroundings and of the possibilities that hitherto laid dormant in it. This last possibility is the way in which the destabilisation in the integration process of allocentric and egocentric frames of reference holds a potential for experiencing a sense of freedom and for establishing a new kind of relationship with space. Of course, the conditions in which one gets lost influence the valence of the disorientation experience. The opening lines of Ray Bradbury’s essay draw this distinction rather elegantly: “To be lost. How frightening. To be safely lost. How wonderful” (p. 61, Bradbury 2017). If we are out at night in a dangerous area of the city, it is likely that our experience of disorientation is strongly negative. If we are late for a meeting and we get lost, we will probably feel helpless and anxious. However, if we are exploring a new city, being lost might be a wonderful way of discovering it, of making acquaintance with its genius loci. And of course, the Situationists decided to get lost in a certain area of the city and had ways of getting back to safety. For them, getting disoriented was a practice, not an accident. This is true also of the Evenki. For them, skilful walking is a practice dependant on a broader culture of navigation geared to a particular environment. Losing our way as a practice is only made possible if we can find our way back.
VIII – CONCLUSION The present analysis of navigational methods, of manakan and of the phenomenology of disorientation (both its negative and positive aspects) advances our understanding of the diversity of wayfinding methods and of the resulting diversity in spatial experience. I have clarified that while it is possible to navigate via the integration of allocentric and egocentric frames of reference, other ways of navigation are possible, as the case of Evenki wayfinding illustrates. I have shown that people using either type of navigational methods can experience disorientation when their process of navigation is impaired, and that their respective experiences (in its negative manifestations) share important phenomenological features. Namely, they both experience body-space reduction and the related anxiety, helplessness and confusion. Finally, I have shown that while the disintegration of egocentric and allocentric frames of reference can lead to this negative manifestation of disorientation, it can also lead to a positive manifestation, and that the phenomenology of this “joyful” disorientation shares the features of the feeling of manakan in Evenki culture. In both cases, the lack of integration of egocentric and allocentric frames of reference leads to a unique relationship with space and to a unique feeling of freedom and possibility. Something that this analysis highlights is that the negative manifestations of disorientation experiences actually comprise two aspects. The first one is the destabilisation in the integration of allocentric and egocentric frames of reference (i.e. integrated self-location), experienced as a destabilisation of the out-of-sight set of spatial objects framing one’s horizon of experience. This is most patent in the feeling that the environment rotates 180º around us when we get “turned around”.
— 53 —
The second aspect is the withdrawal of experienced possibilities that comes with the impairment of our method of navigation. This is most patent in the phenomena of “body space reduction”. What only now becomes clear is that these two aspects can come apart. People who do not rely chiefly on survey knowledge to navigate (e.g. Evenki) can get disoriented, and their phenomenology will resemble the second (body-space reduction) but not the first aspect of the paradigmatic disorientation experience. And at times, people who rely chiefly on survey knowledge to navigate can get disoriented but experience only the first aspect of the paradigmatic disorientation experience, together with a sense of possibility (akin to manakan) instead of with the feeling of body-space reduction. The following table illustrates the analysis: Integrated
It seems that when in the beginning of this chapter we had one central unanswered question, now we find ourselves with several more. We started by asking how it can be that disorientation leads to a sense of possibility. We answered by showing how a lack of integrated self-location can lead to a sense of possibility and in the process gained important phenomenological insight into this connection. However, this insight raises questions about the exact functional story behind this connection. And introducing a double dissociation between body-space reduction and integrated self-location poses the following question: how could we ever characterise disorientation in a way that accounts both for this double dissociation and for the positive aspects of the phenomenon? Chapter 4 will introduce a candidate for such a characterisation. The remaining chapters will build on that characterisation to construct a theory capable of answering these and other questions about the elusive phenomenon that is disorientation.
— 54 —
4 . D I S O R I E N T A T I O N I S A M E T A C O G N I T I V E F E E L I N G
I – METHODOLOGY As we saw in chapter 1, a working characterization of disorientation is a necessary first step in the direction of a unified conceptualization of disorientation. The difficulty is that such a characterization should ideally both spring from the subjective experience of disorientation and be constituted of non-subjective elements. My approach to attain this difficult balance is to tackle head-on the subjective side of disorientation, which is often overlooked in the literature. The methodology in this chapter consists in analyzing different subjective experiences of disorientation from a corpus of cases (hereon referred to as scenarios) that constrain a characterization of disorientation. Scenarios are based on pseudonymized real-life cases that have been gathered through online surveys, semi-structured interviews and direct reports. Scenarios are included insofar as they highlight the relevant dimensions of the phenomenon under study. The corpus is in an early stage, but already advanced enough to provide a good reference for the study of disorientation. As of 29 January 2021, we have received 192 responses. These were complemented by a series of cases of disorientation that were reported directly (e.g. by people that were aware of the existence of the corpus) and then pseudonymized. The survey asked subjects to report on a particular instance in which they felt disoriented. We then collected demographic data from subjects, and we asked them to rate a series of statements about their disorientation experience on a Likert scale (e.g. “the experience of disorientation made the environment feel unfamiliar”). Direct reports were particularly interesting in that the subjects reported about rather unusual cases of disorientation or about cases that were not quite core cases of disorientation but borderline cases (e.g. illusory feelings of orientation and disorientation, discussed in section 2). The development of a corpus of subjective reports is very much in line with work on spatial disorientation in particular and cognitive geography in general (see Montello 2017 for the use of explicit reports in cognitive geography). The idea behind the development of our corpus of cases of disorientation is to have a body of reports that can help us make sense of the phenomenon and guide the conceptual work. I do not intend for the surveys and reports to be direct support for a given hypothesis but to further our general understanding of disorientation. Moreover, I have only chosen cases for this study that I deemed uncontroversial enough to ensure intersubjective agreement. To illustrate this methodology, I will begin by discussing a paradigmatic scenario of disorientation:
Scenario 1. Paradigmatic Disorientation Case. I was trying to look for a bookstore that I had gone to once in a mall. I only remembered the direction from the entrance to that bookstore but didn't remember the route exactly. When I arrived at a junction, I didn't know where to go next. I randomly chose a route and unfortunately it was the wrong way. I tried
— 55 —
to use the interactive map in the mall, but it seemed like the map wasn't properly oriented and it referred to certain landmarks in the mall that I don't know about. I tried to go back to the entrance and find my way again. I succeeded the second time around.
In the above scenario, the subject is objectively lost and feels disoriented. She realizes that she is not in the place that she had previously thought. She chooses the wrong turn and finds herself in an unfamiliar place. She tries to use a map, but she is unable to interpret it. She manages to retrace her steps to the entrance. On the second time around, she manages to find her way to the bookstore and the feeling of disorientation dissipates. My main tenet is that the above paradigmatic scenario – modal as it may be – makes it difficult to disentangle the different aspects of disorientation. To bring into focus the extent of this characterization, in the next section, I differentiate between ‘being lost’ (objective; third-personal) and ‘being disoriented’ (subjective; first-personal).
II – THE MISALIGNMENT BETWEEN BEING LOST AND BEING DISORIENTED From the objective point of view, we can characterize being lost as the subject’s inability to find her way (even if, eventually, the subject regains this ability). This is the characterization that Dudchenko offers in his book, and it does not require access to first personal data (Dudchenko 2010). Whether the subject is able to find her way or not is evident or can be gathered from her behavior, and afterwards, from whether the subject has succeeded or not in finding her way. Alternative objective characterizations of being lost include a failure of the way-finding process (Golledge 1999) or not knowing the directions and distances to get to a given point (Rieser 1999). There is a potential epistemological issue with an objective categorization. Namely, that for determining whether the subject is objectively lost or not, we might need to ask them if they are subjectively disoriented. Nevertheless, the issue is of a practical rather than ontological nature. What matters for the purposes of this section is whether or not the subject is objectively lost, not how an external observer can know whether or not the subject is lost. Furthermore, there are several examples of objective characterizations of being lost that lead to objective measures. For instance, in an early wayfinding study, Best defined the degrees of lostness as the deviations from the most direct route (Best 1970). By their very nature, search and rescue missions also need to use an objective operational characterisation of being lost — the lost person is of course declared lost (objectively, as search and rescue teams cannot ask the lost person directly) before the operation starts. Other studies compare the choices participants make to the optimal choices available to measure lostness (van den Berg 2018) or use a hierarchical Dynamic Bayesian Network model to detect outliers in the subject’s GPS trajectories as potential instances of subjects being lost (Lin et al 2015). Each separate objective measure of lostness has its potential pitfalls (e.g. following the most direct route might not be the most relevant criteria for subjects choosing their itinerary). For this reason, studies often try to find the objective measures of lostness that best predict subjective disorientation based on first-person reports (see Gwizdka and Spence 2007 for an example in the context of lostness in web navigation). Here again, a characterisation of subjective disorientation is a necessary first step to understand the phenomenon that the study in question is trying to capture with different objective measures. An objective characterization of being lost is orthogonal to the subjective characterization of disorientation. This means that a double misalignment can occur: a person can feel disoriented without being lost, and conversely, she can feel oriented when in fact she is lost. Below is a case-
— 56 —
based matrix meant to disentangle the cases in which the objective and the subjective characterizations come apart (numbers refer to the scenarios in the text)14.
Subject feels disoriented Subject does not feel disoriented
Subject is objectively lost Paradigmatic Disorientation Case (1)
Illusory feeling of orientation (2a)
Subject is objectively not lost
Illusory feeling of disorientation (3)
Paradigmatic Orientation Case
A good example of a ‘paradigmatic disorientation case’ is scenario number 1. It is not in the scope of this chapter to analyze orientation cases, but for a paradigmatic orientation case, the reader can pick up from the myriad that one finds in daily life. An everyday activity like going from one’s house to one’s favorite bakery and back is an example of a paradigmatic orientation case. What is of interest here are the other two cases: the occurrence of disorientation without the appropriate feeling, and of orientation with the inappropriate feeling. Scenario number 2 contains an example of the former, and scenario number 3 of the latter.
Scenario 2. Illusory feeling of orientation. “To go to work (29 rue d’Ulm), I come from Rue de l’Estrapade and then through Rue des Irlandais. It is not possible to cycle Rue des Irlandais on the opposite sense, so in the evenings I usually take Rue Amyot [which seems parallel to Rue des Irlandais]. When I started working and I first took Rue d’Amyot on the way back, I felt confused to find myself not in Rue de l’Estrapade, but in Rue Tournefort.”
To make sense of this scenario it is convenient to sub-divide the return path into two sections. Section 2.a extends from rue d’Ulm to Rue Amyot and the length of Rue Amyot before turning into Rue Tournefort, and section 2.b is that of Rue Tournefort (i.e. when the lost person realizes not being in Rue de l’Estrapade and onwards).
14 Daniel Montello makes a similar differentiation between being disoriented (subjective) and being misoriented
(objective) (Montello 2017). The main difference is that Montello characterises subjective disorientation as a belief state and I cast subjective disorientation as an affective state, as will become clear in our next section.
— 57 —
Using a first-person characterization of subjective disorientation, the subject feels disoriented only in section 2.b, when she realizes that she is not in Rue de l’Estrapade. In contrast, if we use a third-person characterization (e.g. deviation from the most direct route), the subject is lost all along the return path, because she is taking a path (rue Amyot) that does not lead to Rue de l’Estrapade, where she wants to go. In other words, the subject is unable to find her way to Rue de l’Estrapade and all of her navigation is a manifestation of this inability. Section 2.b (just like scenario 1) is a case of disorientation for both the third-person and the subjective characterizations. In contrast, there is a misalignment in section 2.a between the two characterizations. Another example of an illusory feeling of orientation is the phenomenon of veering. Souman and colleagues asked subjects to keep a straight path over several hours in a German forest unfamiliar to them (Souman et al 2009). The subjects walking in good weather managed to keep a fairly straight course, but the subjects walking in cloudy conditions started to veer and ended up walking in circles, a common occurrence in lost person behavior (Hill 1998). Of course, part of the problem is that subjects might be under the illusion that they are keeping a straight path when in fact they are veering. Something that should be noted right away is that the feeling of disorientation is not functionally idle. It is the effect of an erroneous or imperfect navigational process, and it can bring about various psychological states and actions, such as anxiety or an urge to act or, conversely, inhibition or poor planning, depending on a variety of factors (Hill, 1998). The feeling in question, disorientation, is a key ingredient for understanding the deployment of further behavior. Moreover, feeling disoriented can be good for you, in particular if you are (objectively) lost. It may inhibit inappropriate behavior, and/or trigger reorientation practices. It may have costs if you are not objectively lost, but these
— 58 —
costs are arguably lesser than the ones incurred in the case in which you are lost without feeling disoriented. A good parallelism to understand the distinction we are after is provided by the distinction between danger and fear. Let us suppose that the formal object of fear is danger (Kenny, 1963). The function of fear is indicating (or registering, or representing) danger, but fear can occur erroneously, when its formal object, danger, is not actually present. John might be afraid while he is walking in the forest at night (because the trees seem to shift shapes or because of the eerie hoots of owls) even if objectively there is no danger. By contrast, if John is happily walking home in the middle of the day in a familiar street, and a mugger is following him to rob him, there is a real danger, but if John is completely unaware of it, he will not feel fear. Fear and danger may thus be not aligned. In the same way, according to my characterization, a subject can feel disoriented without actually being lost. The next scenario makes this vivid.
Scenario 3. Illusory feeling of disorientation. I know my way from my home to my school [29 rue d'Ulm] and from my home to my work [Bastille] very well, but I have had difficulties every time that I have tried to go from rue d'Ulm to Bastille. Yesterday, I was going down rue Claude Bernard to Av. des Gobelins, and I perceived Boulevard Saint Michel as parallel to Claude Bernard [it is actually 135° oblique], which contradicted my mental map. This made me feel disoriented.
I ended up facing Rue du Fer à Moulin, which I vaguely remembered having followed once and that I believed had taken me to Jardin de Plantes. I decided to follow that. I was on the right way, but I still felt disoriented. I felt uncomfortable, I didn't trust my own way-finding, and my mental map was incoherent. When I arrived to Rue Geoffroy Saint-Hillaire I found the path I take every day from my home to work. From then on, my sense of disorientation began to diminish. In scenario 3, the impression that Avenue des Gobelins is parallel to Boulevard Saint Michel contradicted the subject’s internal spatial representation, and disorientation ensued, even if the subject was eventually able to get to his destination (Pont d’Austerlitz), and even if he was following the shortest way throughout. Thus, according to the third-person view, this is not a case of a subject being lost, because she was finally able to find her way to Pont d’Austerlitz. What makes this scenario (and illusory feelings of disorientation in general) so interesting is that although the subject is disoriented, she is still able to carry out the function of finding her way. The illusory feeling of disorientation leads the subject to erroneously assume that he is lost. This stands in contrast to the illusory feeling of orientation, in which the subject erroneously assumes that she is not lost (at the beginning of her route). Of course, the aim of this section is not to completely separate the third-person and subjective aspects of being lost and disorientation. These two are generally concomitant and are only temporarily apart in borderline cases. It is the subjective dimension that defines disorientation in my characterization, but if a subject is unable to find her way (as happens in scenario 2a) the likelihood of the subject experiencing disorientation increases (as is the case by the end of scenario 2b). The opposite is also true; if a subject is on the right way but disorientation occurs (as happens in scenario 3), the likelihood of the subject’s disorientation waning and disappearing is very high (as is the case by the end of scenario 3).
III – THE CORE CLAIM: DISORIENTATION IS A METACOGNITIVE FEELING
— 59 —
If the subjective state of disorientation is not necessarily aligned with either a missing recognition of the place one is in nor with the objective condition of being lost, what does the state consist in? The corpus of disorientation reports that we have collected can help us answer this question. As we saw in chapter 2, the reports show that disorientation has an emotional dimension that should not be neglected. Subjects reported being
• anxious (“I didn't know where I was heading nor to which side of the station I was going. The feeling is horrible and stressful”; “I was anxious really anxious because I didn't want to be late to my first class”),
• helpless and vulnerable (“it made me feel vulnerable”; “I felt unsafe and anxious because no one was with me, and the environment was not familiar to me”; “I just felt confused and helpless — I didn’t know what to do”),
• confused (“I was slightly confused, as all of the streets seemed similar and we kept walking in a circle although we took different turns”)
• and isolated (“I get nervous the longer it takes me to figure out the direction. I attempt to find out which street/avenue I am at. It feels like I am a tiny spec in all this action that’s happening around me”; “Feeling of being left alone, feeling of not being able to make decisions and unable to be independent.”).
Another important affective element in many disorientation episodes is unfamiliarity. 66% of subjects agreed (from somewhat agree to strongly agree in a Likert scale) that the experience of disorientation made their surrounding environment feel unfamiliar. My claim is not that disorientation is always a highly arousing affective experience. There are instances in which disorientation occurs in a low stakes scenario and the agent remains calm throughout. Rather, the arousing instances of disorientation from the corpus help us identify some of the paradigmatic affective elements of the phenomenon. What is important about affective experience in general is that it involves (a) varying levels of valence and arousal and (b) a formal object. Phenomenal valence refers to the aspect of felt positivity (e.g. feeling happy) or negativity (e.g. feeling sad) (Charland 2005; Colombetti 2005) and felt arousal refers to the felt changes in levels of excitement, energy or activation (Colombetti and Harrison 2018). Arousal can be low in some disorientation experiences, but in those cases, there is still a particular valence affectively permeating the experience (negative, in most cases, but see chapter 3). Moreover, as it will soon become clear, there is a formal object of the experience, a way that the experience assesses the situation. As I mentioned in the previous section, another central aspect of disorientation is that it is functionally effective. Roughly speaking, disorientation results from shortcomings in the subject’s active navigational processes and in turn it causes the subject to behave in a variety of ways in order to solve the said shortcomings — disorientation has an evaluative (and a regulative) function. Ideally, the subjective state of disorientation should track the objective state of being lost in order for the subject to stop being lost. The only way to make sense of this evaluative aspect of disorientation is to assume that disorientation is part of a process of metacognition that tracks the performance of active navigational processes. In the prototypical case of disorientation, the metacognitive process would be tracking the performance of the online system of spatial representation15. That is, it would be tracking the success in the integration of allocentric and egocentric frames of reference. Nevertheless, as we saw in chapter 3, there are some cases in which navigation processes do not depend on allocentric-egocentric integration, but on other wayfinding
15 The online system of spatial representation was introduced in chapter 2 in the context of the disorientation
experiments of Waller and Hodgson (Waller and Hodgson, 2006).
— 60 —
methods (e.g. skilful walking in the case of the Evenki). Furthermore, the disorientation often emerges when there is a shortcoming in navigational processes that are distributed beyond the individual. This is an issue that we will go back to in chapter 6, but for the time being, scenario 4 below provides a useful example:
Scenario 4. Cognitive Artefacts. I was in New York City, and I was trying to figure out if I should walk to the left or to the right after exiting a restaurant. The mobile GPS was not well calibrated, and so no matter the direction I walked in, the GPS pointed to the opposite direction.
Here, the subject feels disoriented after his GPS (i.e. not his internal egocentric-allocentric integration) fails. Therefore, in what follows I will refer to active navigational processes, broadly understood, even if in the prototypical case the process that is being tracked is the online system of spatial representation. The main question is how to square these two aspects of disorientation; the emotional aspect and the metacognitive aspect. Here, we can extend the notion of the formal objects of affective states to explain how affective states can be evaluative. Consider once more the relationship between fear and danger. If the formal object of fear is danger, then the affective state fear is evaluating whether or not a particular object (i.e. that to which the formal object is directed at, such as the bear one is afraid of) is dangerous. Furthermore, the fear causes the subject to act in a certain way (e.g. running away from the source of danger). A fundamental aspect of affective states is that they serve an evaluative-regulative function (Proust 2013). The problem is that emotions such as fear seem to track properties that are quite far from disorientation (e.g. danger) and seem to track external objects (e.g. a bear) and not mental processes (e.g. the online system of spatial representation). To be able to account for both the emotional aspect and the metacognitive aspect of disorientation, we need to depart from a conception of affective states as limited to basic emotions such as fear. A different subclass of affective states that has recently attracted a surge in research interest can help us do the necessary explanatory work: metacognitive feelings (see Arango-Muñoz and Michalean 2014 for a review). Metacognitive feelings are described as phenomenal experiences concerning the subject’s own mental states, processes or capacities. Some examples of metacognitive feelings are the feeling of knowing (Koriat, 2000), the tip-of-the-tongue state (Brown and McNeill, 1966), or the feeling of forgetting something (Arango-Muñoz, 2013). The existence of metacognitive feelings provides then a solution to the conundrum of disorientation being both an affective and a metacognitive state: disorientation is a metacognitive affective state. In other words, disorientation is a metacognitive feeling. As I have said, the subjective state of disorientation should track the objective state of being lost, and this can help us precise our characterisation. If we understand, following Dudchenko, being lost as the subject’s inability to find her way, then the metacognitive feeling of disorientation is tracking if the subject is able to find her way. More technically, we refer to disorientation as a metacognitive feeling that evaluates and regulates the subject’s active navigational processes. What I understand by metacognition is the process through which a sub-system evaluates a particular aspect of cognition3. In the standard case of disorientation, what is being evaluated by a metacognitive feeling is the level of confidence in the online system of spatial representation. Among the existing types of metacognitive feelings, disorientation is most akin to the metacognitive feeling of confidence. When feeling disoriented, the subject does not have enough confidence in her capacity to find her way. This low level of confidence is often linked with the
3 The notion of subsytem is clarified in section 4.
— 61 —
other emotional aspects of disorientation such as vulnerability, confusion and anxiety. However, it is this low level of confidence in the subject’s active navigational processes that is at the heart of subjective disorientation. This low level of confidence in disorientation stands in contrast with the feeling of (high, complete) confidence (see Fleming and Dolan, 2012 and Yeung and Summerfield, 2012 for reviews of the feeling of confidence). When a subject feels disoriented, this means that she has a low level of confidence on her active navigational processes (e.g. the online system of spatial representation). Confidence should be understood here as subpersonal-level confidence, not as personal-level confidence (see Dennett, 1969, Hornsby 2000 and Davies 2000 for a discussion of subpersonal and personal levels in psychological explanation). Of course, the two conceptions of confidence tend to correlate, as we can see by subjects feeling unsure so often when disoriented. This is similar to the difference between personal-level surprise (e.g. the surprise a subject experiences in a surprise party) and subpersonal-level surprise (i.e. the surprise generated by stimuli that were unpredicted). As we saw in chapter 2, the online system of spatial representation is dynamic, transient and constantly updated. The same is true of any active navigational process that is being tracked by the metacognitive feeling of disorientation. What I mean by active is that the navigational process is both current and relevant. In other words, the evaluation process is done in a context-sensitive way (Proust, 2015). The process is current in the sense that it is directed at navigating the space in which the subject actually is at a given moment (even when a subject feels disoriented about a remembered situation, the remembered situation is relived as if it were current). What I understand by relevant is that it is an action-oriented process that takes place in a meaningful space. What the relevant space is is determined by the nature of the spatial task. Consider the following scenario:
Scenario 5. Relevance. I had just arrived at NYC and I did not have a map of the city. I needed to go to the 998 Columbus Avenue from midtown: I did so just by following the Columbus Avenue numbering without ever feeling disoriented.
In this case, the subject has a high enough confidence in his active navigational processes, because very little is needed for the spatial task. The necessary elements are his destination (north of Colombus Avenue), the path (998 Colombus Avenue) and his position (updated by checking the street numbers). Even without a good spatial representation of NYC, the subject can follow Colombus Avenue without feeling disoriented, because he can be confident in having these necessary elements that are needed for the spatial task at hand. At the opposite end of the spectrum,
Scenario 6. ‘Offline’ spatial representation. I was in Manhattan last summer for a study-away course. One weekend heading out to hang out with my friends, I decided to take the metro. After arriving at my destination, I attempted to leave the subway station, which required me to take the stairs up to the surface level. I was still able to visualize a map of where I was, uptown and downtown Manhattan, East and West. However, once I was up the stairs and on the sideway, everything got louder, people walking in every direction, sounds of cars, people on their phones... I tried desperately to find any sign of in which direction was what. I got very nervous. It felt like I was a tiny speck in all this action that was happening around me.
Here, the subject has a good offline spatial representation of Manhattan, but because she cannot use it in a relevant way in the space in which she finds herself after coming out of the metro, she feels disoriented. An impersonal map of a remote city, no matter how well known, is not enough for navigational success. What is needed is rather a meaningful representation of the space we are navigating, in which elements such as our destination, possible routes, landmarks, and cardinal directions are embedded in our perception. When the subject doesn't have enough confidence in the process generating this meaningful representation, disorientation occurs.
— 62 —
IV – METHODOLOGY When talking about metacognition as a cognitive subsystem evaluating a given aspect of cognition, there is a tendency to frame metacognition in metarepresentational terms, i.e. in terms of the second-order representation of first-order cognitive content. If this were the case, we would be forced to say that the low confidence that is characteristic of disorientation is a meta-representation, e.g. a representation of a spatial representation. Instead, I will follow Proust’s claim that metacognitive feelings should be characterized in non-metarepresentational (and non-conceptual) terms (Proust, 2007). One can think, consciously, that one is wrong about her path, and even express this thought propositionally (“Alex convinced me that I was in the wrong meeting room”), but one can feel disoriented without thinking that she is and without being able to express it propositionally. In her account, Proust claims that there are several properties (e.g. causal contiguity and epistemic transparency) that are present in metacognition but not in meta-representation. Discussing these properties is outside of the scope of this chapter, but an additional argument for explaining metacognition in non-metarepresentational terms is that doing so is in line with the principle of parsimony: We should not postulate meta-representations if there is a way to explain metacognition without appealing to meta-representations. However, this begs the question of how metacognition works if not through meta-representation. A promising approach is an account of metacognitive feelings according to which somatic cues are ingredients of metacognitive feelings (for an example of this line of research see Koriat and Nussinson, 2009.) A recent experiment in this line of work demonstrated that metamemory beliefs (beliefs about the contents and accuracy of one’s own memories) are positively correlated with interoceptive beliefs and that a subject’s metamemory accuracy is positively correlated with interoceptive accuracy, which indicates that there is a common mechanism subserving both metacognition and interoception (Chua and Bliss-Moreau, 2016). Low confidence in the subject’s active navigational processes would elicit somatic cues that would in turn prompt the metacognitive feeling of disorientation. Of course, it is likely that there are other factors outside somatic cues that contribute to metacognition. For example, in vision-related metacognition, several cues (e.g. evidence for a perceptual decision and visibility of the stimuli) are integrated to generate confidence (Rausch, Hellmann and Zehetleitner, 2018). At the computational level (Marr, 1982), research has successfully used a Bayesian framework to model the production of metacognitive assessments (Fleming and Daw, 2017). The exact mechanisms and cues involved in generating the metacognitive feeling of disorientation remain an open empirical question, which lies outside the scope of this chapter. The important thing for my argument is the plausibility of non-conceptual, non-metarepresentational metacognition, and support for this conception of metacognition might come from work on empirical or somatic cues or work on specific computational models. A more detailed account of how this tracking process works will be given in chapters 5 and 7. The notion of subsystems is worth unpacking. Here, I understand subsystems in functional terms. The idea is that there are various subsystems that track the performance of different aspects of cognition, but at this stage I stay neutral on questions such as the somatic and neural underpinnings of said subsystems. A good example of a functional subsystem is the feeling of familiarity, as understood by the discrepancy-attribution hypothesis. According to Whittlesea and Williams’ discrepancy-attribution hypothesis, the feeling of familiarity arises from the perception of a
— 63 —
discrepancy between the actual and expected fluency of processing (Whittlesea and Williams 1998). See below for an example. If one encounters the clerk from the corner store on a bus, dressed in civilian clothes, one may fail to identify the individual. In that case, one is likely to experience a feeling of familiarity. That feeling is due in part to fluent perception of the clerk's face but that fluency is no greater (probably less) than it was when the individual was met in the store. Instead, the feeling is produced by the discrepancy between fluent processing and a context in which all people should be strangers. -p.560, Whittlesea and Williams, 2000. Although not the only candidate for explaining familiarity, the discrepancy-attribution hypothesis has received substantial empirical support (Whittlesea and Williams, 2001a, 2001b), and it conceptualizes a metacognitive subsystem in functional terms, i.e. as a comparator of expected and actual fluency. In a similar vein, I characterize disorientation in functional terms, as a subsystem that monitors the performance of an active navigational process. There might be a relevant difference between the metacognitive feeling of disorientation and other metacognitive feelings such as the feeling of knowing (e.g. the feeling of knowing what the capital of Peru is even if one is unable to remember it at the moment). In the case of the feeling of knowing, the metacognitive feeling is a ‘comment’ on (i.e. it indicates something about) propositional knowledge. In the case of disorientation, the metacognitive feeling might sometimes be a ‘comment’ on an activity (i.e. the activity of orienting and wayfinding). The following scenario (although not a case of disorientation, but relevant for the general operation of wayfinding) is a case of a metacognitive feeling that is about an action, not about propositional knowledge:
Scenario 7. Feeling the wrongness of an action in space. I started walking just outside Barbizon at the intersection of Allée de Vaches and chemin du Bornage. I noticed that the Chemin du Bornage was almost aligned with the sun. I moved a few hundred meters towards South East, then decided to head South, imagining to keep parallel to Bornage, walking in the woods after having left the path. This meant keeping the Sun a bit in front of me to my right. I noticed that once I had set for a heading, my body tried to keep that heading, no matter what. That is, if I intercepted a track winding right that took me back to Bornage, where I did not want to go, I felt an urge to go left that made me feel uncomfortable as long as I was on the track.
The subject in Scenario 7 reports a tendency to move in a direction that seems to be forced upon him. The subject has the feeling of not doing the right thing. It is a feeling that is associated to the action, not necessarily to a representation of the world; and it is a feeling about the action, not a representation of the action. The feeling in this case is caused by a form of wishful thinking. The subject in the above scenario reports he would like to be able to keep heading South (in the long run) and hopes that the path (that right now is heading West) in the long run will take him South. Feelings associated with actions instead of representations would not fall outside of metacognition; rather, they would be a subset of metacognition. If metacognition is not a meta-representation, but the result of a subsystem tracking another, then a source monitoring an action (such as keeping one’s way) gives rise to a metacognitive feeling (e.g. not doing the right thing) of how one is doing the action. In certain cases, the feeling of not doing the right thing leads to a metacognitive feeling of low confidence in the subject’s active navigational processes, causing the subject to feel disoriented. A question that arises here is if disorientation should not be best understood as a metacognitive feeling indicating low confidence in an action (e.g. wayfinding) or, at the other side of the spectrum, in spatial knowledge, rather than in an active navigational process. The main problem with saying
— 64 —
that disorientation is a metacognitive feeling indicating low confidence in an action is that this characterization is too restrictive, as it would leave outside many scenarios in which disorientation is not a feeling of doing the wrong thing, but of not knowing where one is. On the other hand, characterizing disorientation as a metacognitive feeling tracking knowledge (e.g. “I am at X location”) leaves outside cases like Evenki navigation or cases of navigation distributed beyond the individual such as the one in scenario 4 (cognitive artefacts). In contrast, characterizing disorientation in terms of the subject’s active navigational processes covers all of the above scenarios.
V – EXPLORING THE FACTORS AND EFFECTS OF DISORIENTATION I mentioned above that disorientation feelings, like many other feelings, are not causally and functionally idle. They are not just concomitant factors of cognitive states that monitor one’s location in the environment. Thus, they must have causes and produce consequences. It stands to evolutionary reason that the feeling of disorientation is entangled in a functional web. It is arguably useful to feel disoriented if indeed you are lost, and if it is useful, it is because feeling disoriented can induce a change in behavior that is hopefully conducive to finding one’s way. Causal determinants of the disorientation feeling can coincide with those that induce the subject to get lost objectively or can include other elements. In the first case, I call them straightforward causal determinants. If the light is switched off and you are made to spin, you are likely both to be objectively lost and to feel disoriented. The causal determinants of the feeling of disorientation are here straightforward as they are causal determinants of being lost as well. However, for those cases in which objective being lost and the feeling of disorientation are doubly misaligned, there might be a difference in their respective causes. For instance, in the case explored in scenario 3, Illusory feeling of disorientation, in which the subject erroneously believes to have taken the wrong way when cycling from school to work, there are no elements causing the subject to be lost, but an error in perception (i.e. perceiving oblique streets as parallel) causes the subject to feel disoriented. There is a higher level of complexity in scenario 2, Illusory feeling of orientation. Here, the subject is objectively lost, but doesn't feel disoriented at first. It is only when she ends up in Rue Tournefort instead of in Rue de l’Estrapade that she becomes disoriented. What causes both being lost and subjective disorientation is that the subject had believed that there were four streets forming a square when in fact there were five streets forming a pentagon. What is interesting is that this error causes the subject to be lost right away, while subjective disorientation only arises when the subject arrives to an unexpected street. This shows how tightly linked being lost and feeling disoriented are even in cases of dissociation. The reason is that one of the main roles of the metacognitive feeling of disorientation is to track objectively being lost, so the causes of objectively being lost have a high likelihood of also causing subjective disorientation, even if the inverse might not be the case (as we saw in the example of scenario 3). There are many elements that can play a casual role in the feeling of disorientation. One such element is lack of access to cardinal directions, as seen in scenario 6, ‘Offline’ spatial representation, in which the subject felt disoriented after coming out of the metro in NYC. Other possible elements are erroneous beliefs (e.g. scenario 8 below, in which the incorrect but tenacious belief that the sea is to the South causes the subject to become disoriented), the absence of a mental map (e.g. scenario 9 below, where the subject arrives to a new city and is unable to interpret the
— 65 —
metro map) or the presence of visual impairments (e.g. scenario 10 below, in which the subject is Scooba diving and loses all sense of orientation due to darkness):
Scenario 8. Erroneous belief. Went the wrong way in Malmö, Sweden, due to an irrational belief that the water is to the South when it is in fact to the North. This happened several times.
Scenario 9. Novelty. I found the experience of the Tokyo Subway System very disorienting. I was totally lost. The signs were in a different language and maps were unintelligible.
Scenario 10. Visual Impairments. I was scuba diving in Croatia when we started entering deeper water and it got darker and darker. We then moved under a cliff which made it look as if everything was upside down, leading to many divers making mistakes equalizing their buoyance by inflating the jacket and thus actually moving closer to the cliff. I was not able to rely on my senses and it felt as if I was in a game or a different world of some sort.
Another very interesting, if easily overlooked element, is place recognition (or its absence). The absence of place recognition can elicit disorientation, but interestingly, unexpected place recognition can do so as well. We will look at two real-life scenarios to help us understand the role that place recognition plays in disorientation:
Scenario 11. Place recognition induces the disorientation feeling. In 1990, I first visited Berlin, immediately after the Wall came down. Boundary checks were still enacted. There were only a few passages from West to East. I went through the checks at Checkpoint Charlie, followed a tortuous path, made a turn, saw an “Ausgang” (“Exit”) sign, and exited the Checkpoint – only to find myself once more in the West! The realization that I was back to the starting point was initially of disorientation.
The setting is one of stress and pressure: the person was going for the first time to a country of the former Eastern Europe, in a politically and historically charged context, paid attention to many aspects of the situation (armed soldiers, perceived as threatening) and did not monitor spatial progress or updated his position in an effective way. It is important to note that the disorientation feeling first occurred when he realized that she was back at the starting point. Then, he knew where he was, but he felt disoriented (in fact, it is because of this knowledge that the feeling of disorientation set in.) The disorientation feeling seems to serve here in the monitoring of recent spatial activity, retrospectively (e.g. “I was wrong in my beliefs about my whereabouts” or “I realized that I was in the wrong place, relative to my expectations”). Thus, recognition has the power to induce a disorientation feeling. This becomes even clearer in the following case:
Scenario 12. Fog disorientation: recognition induces disorientation feeling. Coming back from Barbizon, on the Bornage trail near the Maison Forestiere de Macherin. 10am, fog, visibility 40m. I exit the Bornage trail heading towards home, at a precise point we marked with a stone. I know there is no trail and I try to pass a first series of hedges. I keep a bit on the left assuming I will get close to the series of fenced properties one of which is mine. NB I probably came in this area 50 times before. At some point I cross a totally unexpected large trail. I am surprised. I think I am still oriented, heading in the right direction, hence I try to locate that unexpected trail on my mental map. I think: Perhaps it is a new path created by the Bost farmers to collect logs. But all of a sudden, I recognize that it is the Bornage trail: I went full circle, coming back almost to starting point. Now I feel disoriented and have to reset the mental map.
— 66 —
In this scenario, the subject does not feel disoriented at first, even if he is objectively lost. For that reason, when he encounters a trail that he thought he had left behind, he does not recognize it. He did not expect the trail, so he perceives it as a new trail that she did not know exist. Only after the recognition of the trail does the feeling of disorientation arise, followed by a resetting of the subject’s mental map. The “dual” theory of reasoning might help us understand the above scenario and the functional role of the disorientation feeling. According to this framework, there are two modes (also referred to as systems in some of the literature) of operation during decision-making tasks (Evans 2003, 2012, 2015 for reviews; Evans and Frankish 2009; Kahneman 2011). “Mode 1 (M1, for brevity) is an automatic, autonomous, stimulus-driven, fast operating mode that delivers rough but locally acceptable results; M2 [Mode 2] is modulated by will and attention, operates slowly and stepwise, intensely uses working memory, and is in general more accurate” (Casati, 2017). A candidate functional role of the disorientation feeling, as with many other cognitive feelings, is to block a type-1 mode of functioning (M1), and to activate a type-2 mode of functioning (M2)4. The metacognitive feeling of disorientation can shift us from M1 to M2: metacognitive feelings signal that M1 is not functioning, block action, set thought in motion. The person who feels disoriented may stop relying on her intuition about directions and start deploying reflective means of wayfinding (think about the path traveled, look for known landmarks, ask for directions.) Analogously, disorientation might inhibit the use of the online system of spatial representation and facilitate the use of the offline system of spatial representation, in line with Waller and Hodgson’s finding that disorientation decreases target pointing accuracy, but increases inter-object pointing accuracy (i.e. judging the relative directions between different objects in the room, such as “Imagine that you are at the door, facing the TV. Point to the fabric”), which relies on the offline system of spatial representation (see Huan and Allen 2004 for a discussion of dual theories of spatial memory). Before concluding, I would like to mention another relevant factor. Namely, the metacognitive feeling propagation that disorientation can induce: you are disoriented, and this in turn causes you to have a feeling of uncertainty as well. Interestingly, there could even be propagation at the affective level, without actual relevant changes in the cognitive representations or tasks that the metacognitive feelings are monitoring. For instance, if you have an illusory feeling of disorientation (e.g. you suspect you are too close to the marsh, when in reality you aren’t), this could make you feel uncertain about the next steps on the ground, where actually you have no reason to feel so (you are on solid ground), which might in turn propagate to a metacognitive feeling of error (the feeling that there is something wrong with your walking) or even feedback on the feeling of disorientation.
VI – CONCLUSION In chapter 1 had I argued that an adequate characterization of disorientation was essential to link the broad-ranging literature on disorientation into a unified conceptualization of the phenomenon. My strategy in this chapter was to study different scenarios of human disorientation that would constrain its characterization. I showed a misalignment between objective and subjective characterizations (of being lost and of feeling disoriented, respectively). The aim was not to separate being lost and feeling disoriented unreservedly, but to argue that disorientation can be fully characterized subjectively and that the third-personal aspect of being lost is a critical dimension of the phenomenon. I then characterized spatial disorientation as the metacognitive feeling that occurs 4 However, it should be noted that in disorientation episodes of high stress, the subject might be locked in Mode 1, as
stress is “a state in which an individual is unable to instigate a clear pattern of behavior to remove or alter the event (…) that is threatening an existing goal” (pp. 206–207, Power & Dalgleish, 1997).
— 67 —
when the subject has a critical low-level confidence in her active navigational processes. In the final section of this chapter, I provided a link between the causes of disorientation, the subjective experience of disorientation, and the resulting behavior. The central claim of this chapter is also the central claim of the monograph: disorientation is a metacognitive feeling that evaluates and regulates the subject’s active navigational processes. To develop this point, we need a more precise account of how affective states evaluate and regulate cognition. Chapters 5 and 6 will provide such an account within the Predictive Processing framework, and chapter 7 will do likewise within the complimentary framework of Distributed Cognition. The resulting synthesis in chapter 8 will serve to develop a full-blown theory of disorientation.
— 68 —
5 . I N S E A R C H O F A F F E C T I V E E X P E R I E N C E I N T H E P R E D I C T I V E M I N D
I – INTRODUCTION In the previous chapter, I characterised disorientation as a metacognitive feeling of low confidence in the subject’s active navigational process. To fully understand the way in which the feeling of disorientation evaluates and regulates navigation, we need an account of the evaluative and regulative role of affective experiences. In the next two chapters, I aim to develop such an account with the help of Predictive Processing (henceforth, PP), a novel framework that aims to provide a unifying vision of human cognition. In section 5.2, I outline a set of commonalities of affective experiences, thereby introducing the explananda that a PP account of affective experiences should aim to explain: affective experiences are phenomenally conscious and valenced, they motivate behaviour, and they are intentional states with a particular and a formal object. After introducing PP in section 5.3, I will look at different theories that have been proposed within the PP framework to account for affective experiences. I divide these into two families of theories: Interoceptive Inference Theories and Error Dynamics Theories. In section 5.4, I discuss Interoceptive Inference Theories and argue that while they provide a good mechanism within which to understand affective processes, they face some challenges when it comes to accounting for what I deem the mark of the affective, i.e. that which distinguishes affective states from other classes of states and processes (e.g. non-affective bodily sensations). I posit that a computational account of valence might be able to address this challenge in a way that remains compatible with Interoceptive Inference Theories. In section 5.5, I discuss Error Dynamics Theories and argue that while they manage to account for the mark of the affective through their computational conceptualisation of valence, they are incompatible with one of the central aspects of affective experiences: that affective experiences are phenomenally conscious. The next chapter will build on this review in order to provide a synthesis of Interoceptive Inference Theories and Error Dynamics Theories, in an effort to account for the commonalities of affective experiences outlined in the second section of the present chapter.
II – AFFECTIVE EXPERIENCES What does a tired toddler, a relaxed jazz player, a lascivious llama, a scared cat, a hopeful rebel, a confused voter and a confident nurse have in common? They all are in the grip of an affective experience — assuming that they feel their respective states. Affective experiences or feelings16 are a rich class of phenomena. Nevertheless, we find significant commonalities that affective experiences share among themselves and not with other mental states, distinguishing them as a class. These commonalities constitute the explananda that a good theory of affective experiences needs to account for. In the present section, I will start by outlining these commonalities, which run through different strands of philosophical and empirical work on affective states. 16 I use the terms “affective experiences” and “feelings” interchangeably.
— 69 —
First off, feelings are conscious in that they are felt.17 They are conscious states in the same way perceptual experiences are conscious states. I seek to make this clear by referring to feelings also as affective experiences.18 Moreover, feelings are phenomenally conscious, i.e. there is “something it is like” to have an affective experience. Feelings, in particular, have an affective phenomenal nature. A central aspect of this nature is phenomenal valence (Barrett and Bliss-Moreau 2009). Phenomenal valence refers to the aspect of felt positivity (e.g. feeling relaxed) or negativity (e.g. feeling afraid) (Charland 2005; Colombetti 2005) and is often understood in algedonic terms such as (un)pleasantness or evaluative terms such as seeming (dis)value (Carruthers 2017; Teroni 2018). It is worth emphasizing that when I speak of valence here, I mean to refer to valence as a phenomenal property, i.e. to felt or experienced valence.19 Valence is often regarded as the mark of the affective (Charland 2005; Barrett 2006), that is, the feature that distinguishes characteristically affective phenomena from other non-affective phenomena. This conception of the mark of the affective echoes the conception of the mark (or marks) of the mental, the feature (or set of features) “that set characteristically mental phenomena apart from the characteristically physical phenomena” (Pernu 2017, p. 1). Phenomenal valence plausibly grounds the evaluative dimension characteristic of affective experiences (e.g. Helm 2009; Bain 2013; Deonna and Teroni 2017). For the relaxed jazz player and the hopeful rebel mentioned earlier something is going well (i.e. they are undergoing a positively valenced experience) while there is something amiss for the scared cat and the confused voter (i.e. they are undergoing a negatively valenced experience). Another important phenomenal dimension of affective experience is felt arousal. Felt arousal describes the felt and more or less localized decrease or increase (i.e. change) in level of activation, energy or excitement (Proust 2015; Colombetti 2018). Such felt arousal correlates with actual physiological arousal states that are functionally associated with resource allocation (Dawson et al., 1989; Filion, Dawson, Schell, & Hazlett, 1991). As way of an example, the mentioned relaxed jazz player would feel positive valence and a decrease in her level of excitement. In contrast, the scared cat would undergo an affective state with negative valence and high arousal. It is worth emphasizing that when I speak of valence and arousal here, I mean to refer to valence and arousal as phenomenal properties, i.e. to felt or experienced valence or arousal. These qualities often but not
17 It is, thus, not enough to just point to states which have a similar impact as feelings but are not felt (Lacewing 2007,
p. 97; see also Winkielman et al. 2005).
18 It seems largely a matter of definition that feelings are conscious: “That there can be no unconscious feelings is still the position of “commonsense.”” (Lacewing 2007, p. 98; see also Clore 1994) However, this is not to say that there are no unconscious emotions or, more generally, unconscious states that are functionally largely analogues to feelings or sub-classes of affective experiences like emotional feelings (Winkielman and Berridge 2004; Lacewing 2007). For perceptual experiences it seems plausible that there are forms of unconscious perception and something similar might be said for affective experiences (say, there might be unconscious affective reactions) (Prinz 2005). However, when making this comparison the aspect of interest is not that perceptual experiences are a form of perception (as is unconscious perception) but that they are a form of experience, and it is a conceptual fact about experiences that they are conscious. Affective experiences should also be distinguished from emotional or affective processing which often does occur unconsciously (see e.g. Mathews and MacLeod 2002). Apart from terminological rationale, there is good reason to believe that fine-grained affective experiences require conscious awareness (Pessoa 2005).
19 This phenomenal quality often but not always correlates with closely associated but ultimately non-phenomenal properties such as object valence (Colombetti 2005). It is also worth mentioning that the idea of “unconscious valence” is lately gaining ground (e.g. Berridge and Kringelbach 2015). Unconscious valence has a functional profile similar to phenomenal valence in motivating behaviour. Although I think there might be some problems with this idea, arguing the point would take us too far afield. Here, my focus are affective experiences and I are thus concerned with phenomenal valence. As I will see, however, a virtue of the emerging picture is that it can incorporate the idea of unconscious valence.
— 70 —
always correlate with closely associated but ultimately non-phenomenal properties such as object valence or physiological arousal (Colombetti 2005; 2018).20 This evaluative dimension, in turn, is closely associated with another important feature of affective experiences: they are motivational, in that they exert a motivational push and directly motivate behaviour and action (Corns 2014; Scarantino 2014; Carruthers 2017; Kozuch 2018). Generally speaking, the kinds of behaviours that feelings motivate can be described as aversive (cessation, avoidance) or appetitive (continuation, approach) in nature (Corns 2014). And while it is at times not obvious which specific behaviours some feelings mandate, often the motivated behaviours are quite specific. The motivational force of feelings also reaches into the future: they do not only motivate actions in the moment of their occurrence but make certain behaviours more or less likely in the future. For instance, if you stand close to a fire you might experience pain which will not only motivate you to recoil in the moment but will also make it less likely that you will stand that close to a comparable fire in the future (Kozuch 2018). The motivational force and nature of the motivated behaviour is in part determined by the valence of the experienced affective state. This is underlined by the fact that the polarity and intensity of valence usually map onto corresponding dimensions of the motivational push: strongly/weakly (valence intensity) positive/negative (valence polarity) feelings tend to exert a strong/weak motivational push (motivation intensity) towards appetitive/aversive behaviours (motivation polarity) (Kozuch 2018). Although there is a clear link between valence and the motivation of behaviour, one should be wary of making this link too tight. There are feelings that do not seem to display this tight connection. Affective states such as guilt could appear as negatively valenced as fear (i.e. a highly motivating feeling) but have a weaker link to specific behaviour (Schroeder 2004). We find more striking examples in the context of addiction, where valence and motivation sometimes dissociate (Corns 2014). A heavy smoker regularly has cigarettes that feel unpleasant, but they are highly motivated to smoke despite the affective experience of smoking being negatively valenced. Note that it is not only that the (weakly) negative valence does not lead to a weakly negative motivation to avoid cigarettes but also that there is a strong, positive motivation in place that is poorly explained by the implicated negative valence. Nevertheless, the motivational nature of feelings is something that a good theory of affective states must be able to account for. If valence brings to the fore the evaluative dimension of affectivity, the link between affective states and the motivation of behaviour brings to the fore its regulative dimension. Finally, feelings are not only phenomenal but also intentional states (e.g. Tye 2008; Goldie 2002; Kriegel 2014). They are “about” or are “directed at” something. Feeling pain in one’s wrist is about (the bodily events in) one’s wrist, being afraid of William James’ famous bear is about the bear. The individual objects feelings are directed at are traditionally called the intentional or particular object of a feeling. Crucially, affective experiences cast their objects in a specific light. The pain is not only about one’s wrist but also about its being unpleasant. The fear is not only about the bear but also about its being dangerous. The feeling, plausibly in part via its valence, signals a feeling-specific property exhibited by its particular object. This property specific to a feeling is traditionally called its formal object or core relational theme, which I foreshadowed in the chapter 4 (Kenny 1963; de Sousa 1987; Lazarus 1991; Prinz 2004). To summarise: a feeling represents its particular object as having a feeling-specific property, its formal object.
20 Object valence refers to whether an object or stimulus is negative (e.g. angry or sad faces, snakes) or positive (e.g.
happy and attractive faces, tasty food) as such —objectively as it were—, regardless of whether it elicits emotions or feelings. Physiological arousal are bodily changes (e.g. changes in heart-rate, blood pressure, gland activity, posture, muscle tension) that prepare the organism for response. These changes are sometimes consciously represented as felt arousal.
— 71 —
Note that valence emerges as a property of affective experiences that does not only intuitively distinguish them from other kinds of mental states but also pulls their other features naturally together: First, it (partially) grounds the way in which feelings are evaluative, supplying a general positivity or negativity which is then further specified in the form of feeling-specific formal objects. Second, valence seems (in part) responsible for why affective experiences motivate behaviour and thus can fulfil their regulative function. These considerations make valence appear as an excellent candidate for the mark of the affective. Before introducing PP in the next section, let us recap the features that distinguish affective experiences and that constitute the explananda for any potential explanans within PP:
• Affective experiences are phenomenally conscious, • Affective experiences are valenced, • Affective experience vary in their level of arousal, • Affective experiences motivate behaviour, • Affective experiences are intentional states with a particular and a formal object.
III – PREDICTIVE PROCESSING The main idea behind the PP framework is that the brain is a dynamical, hierarchical, Bayesian hypothesis-testing mechanism. Based on prior information, the brain actively generates predictions about the world by using an internal model. These predictions go top-down (i.e. from abstract levels down to sensory organs and effectors) and side-ways (i.e. laterally across a given level). In turn, the errors of these predictions (prediction error, henceforth Error) go side-ways and bottom-up and are used to update the system’s predictions in a continuous feedback loop. The overall goal of the system is to minimise Error over time (for detailed accounts defending PP, see Clark 2013 and Hohwy 2013). Perception is then conceived as the process of continuously explaining away Error through ongoing predictions in order to successfully represent the world, understood as the hidden causes of changes in sensory input. Each layer in the hierarchy tries to predict the input of the layer below, using models developed to capture regularities in the variation of sensory signals. Sensory receptors capture the mismatch between this cascade of predictions and incoming sensory input (i.e. Error) and the Error that cannot be explained away solely by lower layers adapting their predictions travels upward in the hierarchy, through forward error signals (Friston 2003). Error is reduced over time both by making better predictions about the world and by acting on the world to, on the one hand, acquire (i.e. sample) better information, and on the other, transform it in order to fulfil the predictions (Friston et al. 2011; FitzGerald et al. 2012). Against this background, an action happens in order to fulfil an emergent proprioceptive prediction. For example, in order for an agent to open a door, the system predicts (at a high level) the required movement to open the door and a cascade of lower level predictions ensues (e.g. predictions about the movement of the hand, the fingers, etc.) If the predictions about the movement of the hand or the fingers did not occur, a considerable increase in Error would arise. The movement of the agent, then, happens to minimise the Error of the counterfactual proprioceptive predictions corresponding to the opening of the door. The idea that action occurs to fulfil (proprioceptive) predictions implies that both action and perception are part of the same process of minimising Error over time, commonly referred to as active inference (Friston et al. 2016). A last way in which action and perception are intertwined is through the epistemic role of action. Actions can serve to gather new information, and thus to support perceptual hypothesis. The ways in which action fulfils this epistemic role can be go from
— 72 —
moving closer to a target of interest to the saccadic movement of the eyes (Friston, Adams, et al., 2012). The system uses Error to constantly update its predictions, but not all Error is equally reliable. For example, the Error coming from stimuli with a high level of noise (e.g. a poorly lit room) is less reliable than the Error coming from stimuli with a low level of noise (e.g. a well-lit room), because the Error of the former is more likely (relatively) to be due to noise than to the inaccuracy of the current prediction. Thus, not all Error has the same weight when it comes to updating hypotheses21 about the world. The errors coming from sources that are expected to have a high variance are assigned a lower weight compared to the errors coming from sources that are expected to have a low variance. The inverse of variance is called precision (which quantifies the degree of certainty about the signals; see Feldman and Friston 2010 for a free-energy formulation), and the brain, in addition to first-order predictions, is constantly trying to predict the precision of forthcoming Error. Thus, when the system is choosing among a variety of hypotheses, it is not only weighting how much Error different hypotheses are expected to generate, but also the expected precision that each hypothesis is expected to generate. Accordingly, the brain is always trying to optimise precision, mainly by sampling the stimuli that are predicted to have high precision, which is conducive to Error minimisation over time. In the PP literature, this process of precision-optimisation is what defines attention (Hohwy 2012). PP is particularly compelling in explaining the workings of attention and how attention relates to consciousness. Hohwy suggests that “conscious perception is determined by the prediction or hypothesis with the highest overall posterior probability” (Hohwy 2012, p. 3). The posterior probability quantifies how probable the prediction is given the observed evidence (i.e. the actual sensory input). In other words, the brain is continuously computing and evaluating a plethora of hypotheses about the world, and the hypothesis that is deemed to be the most probable (i.e. the most likely state of the world) determines what the subject is consciously experiencing. Attention has a strong effect on the content of consciousness because it biases the competition between different hypotheses (on account of the estimated precision that each is expected to generate), so that whichever is the most probable hypothesis determines conscious experience. This ongoing competition between different hypotheses is in line with the idea of affordance competition (Cisek 2007). The affordance competition hypothesis emerges as a response to the ‘passive reconstruction’ view in which the brain first creates a unified, detailed action-neutral representation of the world and then this representation is used for planning action. Cisek and Kasala take, on the one hand, the repeated failure to find such action-neutral inner representations and, on the other, the increasing body of neuroscientific research pointing to information being computed in parallel and only integrated when action requires it, to support the idea that potential actions (i.e. affordances) are computed in parallel and compete against each other for further processing. According to their computational model, information is collected to bias this competition until a single affordance is selected for action (Cisek and Kasala 2010). Although this notion originated focusing on action, Andy Clark reinterprets affordance competition in the light of the PP framework and argues that “the brain is constantly computing […] a large set of possible actions and that such partial, parallel, ongoing computations involve neural encodings that fail to respect familiar distinctions between perceiving, cognizing, and acting.” (Clark 2015, p. 177). This last part is particularly interesting because it points to affordances (i.e. action opportunities) being computed in a similar way (i.e. in the computational terms of prediction, precision and error) to perceptual hypotheses.
21 Although the terms prediction and hypothesis are sometimes used interchangeably in the PP literature, technically, a
hypothesis is a joint set of predictions. Inference is the updating of the prior to the posterior in the light of the prediction and the prediction error.
— 73 —
The affordance competition hypothesis hints at a conception of PP in which representations in PP become action-oriented representations: action (i.e. affordance selection) guides the system in processing information further, thus making more detailed predictions about the environment that in turn guide further action. Computational frugality completes this radical conception of PP. Instead of “mirroring” the world through a rich reconstruction, the PP organism goes for action-oriented models that are as simple (and as accurate, of course) as possible: “Bayes optimal agents seek both to maximize the accuracy of their predictions and to minimize the complexity of the models they use to generate those predictions” (Fitzgerald et al 2014, p.1). The intuitive idea here is that the system will try to be efficient with the resources used in reducing Error over time. This drive for efficiency implies distributing the work between brain, body and environment, which involves cultural practices that increase predictability (see Hutchins 2014 for an example of how queuing achieves this through dimensionality reduction), the exploitation of the passive dynamics of the body (the way that bipedal motion relies on the organisation and kinematics of the legs; McGeer 1990) and a large number of heuristics. A good example of the latter is how baseball players (instead of calculating the future trajectory of the ball passively) dynamically position themselves in a way that makes the ball seem to move at a constant speed through the visual field, a trick known as Optical Acceleration Cancelation (OAC). Put in the PP terms of precision weighting, OAC consists in ascribing high gain to the Error associated with cancelling the vertical acceleration of the ball’s optical projection and quashing other errors (Clark 2015). Through precision-weighting, the PP system manages to flexibly switch between strategies that minimise the work for the brain by distributing it instead across the body and the world.
IV – INTEROCEPTIVE INFERENCE THEORIES OF AFFECTIVE EXPERIENCE Within PP two families of theories of affective experience have been proposed so far: the first can be called “Interoceptive Inference Theories” (IIT) (e.g. Hohwy 2011, Seth et al. 2012; Seth 2015; Seth and Friston 2016; Gu et al. 2013; Barrett and Simmons 2015). The second I call “Error Dynamics Theories” (EDT) (Van de Cruys and Wagemans 2011; Van de Cruys 2017 Jofilly and Coricelli 2013). I will discuss both in turn. IIT takes as its point of departure the fact that the central imperative of a biological system is to ensure its survival. Arguably, the most immediate determinant of survival is one’s ability to achieve and maintain physiological integrity or homeostasis, i.e. to keep the body’s vital parameters within viable bounds (Seth 2015). Consequently, all PP is “ultimately in the service of maintaining organismic homeostasis” (Seth 2015, p. 9; see also Barrett 2017). The activity of achieving homeostasis through physiological and behavioural change, is called allostasis. Interoception, the sense of the internal physiological condition of the body, is linked particularly closely to allostasis because it monitors homeostatic change. Against this background, IIT was first developed as a PP-theory of interoception and then taken to naturally extend to affective experiences. It is worth noting that the idea that interoception is a key determinant of emotion can be traced back to at least Aristotle and the Stoics (Aristotle 1986; Gill 2010) and has been famously developed by James (1884) and Lange (1887) into an influential view that counts prominent recent proponents (Damasio 1994; Damasio and Carvalho 2013; Prinz 2004, 2006; Craig 2003, 2009; Critchley et al. 2004). What IIT brings to the table is the idea that interoception is not a feedforward process (as e.g. in Craig 2003, 2009), but part of the larger PP machinery. The initial idea of IIT was that interoceptive experiences – among which we also find affective experiences according to IIT – result from inferences to the causes of interoceptive signals. In Seth’s own words:
— 74 —
On this theory of interoceptive inference […], emotional states (i.e., subjective feeling states) arise from top-down predictive inference of the causes of interoceptive sensory signals […] In direct analogy to exteroceptive PP, emotional content is constitutively specified by the content of top-down interoceptive predictions at a given time.
-Seth 2015, p. 9
Note that against this background, changes in interoceptive signals will lead to Error that will have to be accounted for by adjusting one’s predictions in order to accommodate the Error-generating changes in interoceptive signals. There is something appealing about this idea: it offers a neat account for the intentionality of affective experiences in analogy with exteroceptive experiences. As I have outlined before: affective experiences are about particular objects. If one becomes afraid of an approaching bear, it is the bear that is the particular object of one’s fear. Here one might initially worry that by appealing to interoceptive signals, IIT risks to be overly “body-centric”, a criticism often levied against non-PP versions of somatic representationalism about affective experiences (e.g. as found in Tye 1995). However, the PP framework in which IIT operates can easily rise to this challenge: It is the causes of interoceptive signals that interoceptive inferences represent. These can be bodily (or interoceptive) events as in the case of hunger or pain. However, not only bodily events influence interoceptive signals but plausibly also events or things in the world (or the mind), such as (imagined) bears. One might now wonder: How do non-bodily events get to change interoceptive signals so as to be inferred as their causes in the first place? Low blood glucose levels or physiological damage are clearly integrated within the interoceptive system – but bears don’t seem to be. Before we turn to this question of (extrasomatic) distal causation of affective experiences, it is helpful to consider another question that will bring IIT’s answer in clear relief. Here, we’re after the mark of the affective in the predictive mind: we are looking for something that demarcates affective from non-affective experiences within the PP machinery. Outside of PP-theorising, valence seems the most promising candidate for the mark of the affective. Thus, what we are looking for is an account of valence in IIT terms. Now, is there a mark of the affective according to IIT? Or: is there a plausible idea of what valence could be? And if yes, what is it? What distinguishes affective states from non-affective states in IIT, if anything? Obviously, the mark of the affective cannot simply be that we are engaging in predictive causal inference – since this kind of inference also undergirds e.g. visual experiences that are intuitively non-affective. The remaining candidate is that such inference is interoceptive rather than exteroceptive. In fact, this seems to be what proponents of IIT have in mind: “Emotion or affective content […] becomes an attribute of any representation that generates interoceptive predictions” (Seth and Friston 2016, p. 5). This indeed distinguishes interoceptive predictions from e.g. visual predictions. Thus, IIT’s mark of the affective is the property of certain predictions being interoceptive. So as soon as there are predictions about the causes of interoceptive signals, I experience affective experiences, comprising bodily and emotional feelings. States based on interoceptive inference, thus, will have a valence. Now, it is true that affective experiences often appear to have a bodily component, just think of bodily pain and orgasms. The problem with this proposal, however, is that not everything that has a bodily component seems to be an affective experience. That is, we seem to have experiences that are plausibly construed as interoceptive and grounded in interoceptive inference but that are not affective. Examples are feeling one’s heartbeat, stomach at work or breathing in and out. It appears plausible to draw a distinction between such bodily sensations and bodily feelings such as the mentioned bodily pains and orgasms. One can thus distinguish between cases of “hot” affective
— 75 —
interoception (i.e. bodily feelings) and “cold” non-affective interoception (i.e. bodily sensations) (cf. Proust 2015, p. 20 sq.; Gerrans 2015). The need for such a distinction is suggested by a popular view of bodily pain, which decomposes it into a sensory-discriminative and an affective-motivational component (Melzack and Wall 1988; Auvray et al. 2010). Moreover, we observe dissociations between these components (e.g. Rubins and Friedman 1948; Berthier et al. 1988; Corder et al. 2019). In pain asymbolia, for instance, subjects show a striking indifference to pain while appearing fully aware of the pain, or at least of its sensory-discriminative aspect (Bain 2014; Klein 2015). On the other hand, pain affect without pain sensation has been documented in humans and rats (Ploner et al. 1999; Uhelski et al. 2012). The Cotard syndrome generalises this lesson to other affective experiences. It is known for producing the Cotard delusion where patients come to believe that they are dead. Something in the experience of Cotard patients gives rise and explains this delusion: Cotard patients display a general loss of affect, without impairment in their interoceptive awareness (Michal et al. 2014). In other words, Cotard patients’ bodily sensations are intact but their conscious life has been purged of affective experiences (Gerrans 2015). The emerging issue for IIT’s initial mark of the affective is that it classifies states as affective that aren’t plausibly so. Although interoceptive predictions naturally comprise affective experiences, they do also comprise non-affective experiences. However, recently IIT versions have been developed in a way that bears promise in accounting for the outlined challenge (see especially Seth and Tsakiris 2018; but also Seth 2015). The problem for the initially sketched version of IIT was that it gave us little guidance to distinguish between different kinds of interoceptive inference. Now Seth and Tsakiris’ (S&T) development provides exactly this: generally, they distinguish between probabilistic inference and active inference. Probabilistic inference is inference to the (hidden) causes of sensory signals. It serves Error minimisation by updating predictions based on incoming Error. Active inference, on the other hand, minimises Error by means of issuing actions. Importantly, it can do so in two ways: actions might be directed at selectively sampling information to enhance the models of the predictive system. S&T call this kind of active inference epistemic (active) inference. Then there is what they call instrumental (active) inference. In instrumental inference, actions are performed to exert predictive control over sensory variables in order to bring them in line with the system’s predictions. These conceptual resources can now be put to work to distinguish between different kinds of interoceptive inference. S&T specifically point towards instrumental interoceptive inference which, rather than inferring the causes of interoceptive signals (which could be called “causal interoceptive inference”), is in the business of model-based regulation and control of interoceptive variables (notably, physiological essential variables). In other words, instrumental interoceptive inference directly brings about physiological homeostasis and thus implements the most immediate kind of allostasis. It does so by issuing initially counterfactual interoceptive predictions that result in the engagement of autonomic reflex arcs (“intero-actions”) and potentially other allostatic actions which, in turn, bring about the ultimately self-fulfilling predictions. S&T suggest that instrumental interoceptive inference plausibly comprises emotional (and other self-related) experiences:
[It] is not enough to say that emotional and self-related experiences are the way they are (and are different to, for example, visual experiences) because they emphasise predictions about interoceptive (rather than visual) signals. Instead, it is helpful to consider the nature of
— 76 —
predictions associated with interoceptive inference, especially their control-oriented (instrumental) bias.
-S&T, p. 6-7 In this vein, one might try to draw the initially problematic distinction between hot and cold interoception precisely along the lines of instrumental interoceptive inference and other kinds of interoceptive inference. Importantly, we thereby identify a potential candidate for a refined mark of the affective according to IIT: it is not interoceptive inference simpliciter that makes an experience affective but specifically instrumental interoceptive inference. If there are interoceptive predictions geared towards regulating the internal milieu, we experience affect. Now we are also able to answer the question of extrabodily distal causation of affective experiences: How do non-bodily events get to change interoceptive signals so as to be inferred as their causes in the first place? Think about the fear of the approaching bear again. Spotting a bear will, among other things, plausibly lead to instrumental interoceptive inference (your models of a bear encounter mandate that you better prepare your body for the occasion) which, in turn, will lead to a change in interoceptive signals. Causal interoceptive and non-interoceptive inference will now infer the bear (together with proximal interoceptive factors recruited by instrumental interoceptive inference) as the cause of the (changes in) interoceptive signals. This is a story that illustrates the strength of the emerging picture. However, an initial complication with instrumental interoceptive inference as the mark of the affective seems to be that S&T themselves subsume seemingly non-affective experiences under instrumental interoceptive inference as well. In fact, the explicit target of the article in which they develop the notion of instrumental interoceptive inference is not affective experiences but self-related experiences or “experiences of embodied selfhood” within which S&T count: “experiences of being an embodied organism, experiences of mood and emotion, pre-reflective experiences of selfhood and ‘mineness’, explicit self-awareness, metacognitive insight, reflective self-awareness, and social aspects of selfhood” (S&T, p. 7). It is far from obvious that all these experiences are affective. According to S&T, however, something unites these experiences: they are all “grounded in processes of instrumental (control-oriented) interoceptive inference that underpin allostatic regulation of physiological essential variables” (S&T, p. 10). It seems, thus, that instrumental interoceptive inference is not as much intended as a mark of the affective than as a mark of self-relatedness. Apart from this complication which could potentially be resolved (by e.g. arguing that all the mentioned experiences are, in fact, affective), we run into other problems when applying instrumental interoceptive inference as the mark of the affective. A challenge is posed by the fact that we often have affective experiences where there seems to be plausibly no need for predictive interoceptive control. We find such affective experiences for instance in the context of aesthetics. Consider an example that Jennifer Corns recently advanced against the IIT-congenial idea of valence —what she calls hedonic episodes or (dis)pleasure— as homeostatic utility or disutility:
A feature is homeostatically controlled only if it is continually monitored and subject to stabilizing processes when errors are detected. It is implausible that every stimulus capable of causing a hedonic episode exhibits features subject to homeostatic control. Implausible, that is, that (dis)pleasure is exhausted by homeostatic (dis)utility. Consider the hedonic episode caused by listening to your favourite piece of music. Is it at all plausible that I have a homeostatic mechanism that continuously monitors my Chopin Cello Sonata in G minor levels? Is it any more plausible that I have a homeostatic mechanism that continuously monitors my music levels? How could such a mechanism have evolved and how could it be realized? What are the detectors constantly monitoring music levels? What are the built-in range of values? Counterexamples proliferate.
— 77 —
-Corns 2014, 241
In analogy: what kind of (control-oriented) interoceptive predictions could events such as listening to Chopin, watching a sunset or hearing a joke encompass? It is not easy to think of any. IIT might be able to counter: the interoceptive predictions in question are not to be understood as resulting in control of interoceptive variables right now but later. Note that the intuitive idea behind instrumental interoceptive inference shifts significantly as soon as the counterfactual interoceptive predictions are not about now but about later. These counterfactual interoceptive predictions cannot straightforwardly be called control-oriented anymore, since they are not issuing interoceptive changes (now) but rather predict interoceptive changes in the future. Consequently, what IIT would need to say is that the organism infers that listening to Chopin now will lead to interoceptive changes later. But then: what are the predicted interoceptive changes associated with Chopin? It is still challenging to think of any. Note further that this reply would complicate matters for IIT in trying to account for the host of bodily sensations that I sometimes have in the very moment of undergoing aesthetic affective experiences (think of goosebumps). Since IIT must claim that the instrumental interoceptive inferences concern interoceptive changes in the future, the fact that there are interoceptive changes right now seems not to be predicted by IIT. This is for the simple reason that there is 1) apparently no direct track for music to affect interoceptive signals, and 2) music does not appear to mandate instantaneous regulation of interoceptive signals — what priors might make it seem like a good idea to tinker with essential physiological variables in response to music? Thus, it seems to be an extra-theoretical fact to IIT that in such situations there are bodily events going on. Even if IIT could come up with stories here, there would still be another major problem: affective experiences and their valence have polarity and intensity (cf. Kozuch 2018). Sometimes an affective experience is positive and sometimes it is negative. Sometimes an affective experience is more intensely positive (or negative) and sometimes less. It is highly plausible that these dimensions should covary with properties of interoceptive predictions. But it is not obvious how these dimensions of valence map onto dimensions of (instrumental) interoceptive predictions. Looking at polarity, a straightforward candidate might come to mind: On the one hand, an affective experience is positive if the predicted interoceptive changes are conducive to homeostasis. On the other hand, an affective experience is negative if the predicted interoceptive changes are unfavourable to homeostasis. The problem with this idea is that it includes elements that are not elaborated by IIT: on top of predictions of interoceptive changes (something IIT supplies) there is an evaluation of these changes (something that IIT does not supply). In what way are we to understand this evaluative step? In terms of Error? In terms of precision? It is unclear, partly because Error and precision are nothing specific to interoceptive inference — is it specifically Error occurring in, or expected precision of, interoceptive inference? Instead of explaining the (plausibly valence-based) evaluation, IIT seems to be forced to posit it without being able to account for it. I will return to a suggestion for explaining the mentioned evaluation in section 5.4. Note further that there is a general issue with such “interoception-only” moves: giving interoceptive inference exclusive rights when it comes to affect production appears in tension with one of IIT’s main premises saying that all (not only interoceptive) PP activity is geared towards promoting allostasis:
Interoceptive inference […] should not be considered as a generalisation of predictive coding from exteroceptive modalities such as vision. Instead, perceptual content in all modalities, including modalities such as vision, is a consequence or generalisation of a fundamental imperative towards physiological regulation. Seen this way, all perceptual
— 78 —
content is underpinned by inferential mechanisms that have a functional, ontological, and phylogenetic basis in allostasis.
-S&T, 11
While it is plausible that interoceptive inference is somewhat affectively privileged due to its proximity to homeostasis, it appears unwarranted to grant it exclusive affect-production rights. Let us turn now to valence intensity. A possibility is that intensity varies depending on how far in the future the interoceptive changes are predicted. The further away, the less intense, and the more imminent, the more intense. This can’t be the whole story, though: the IIT story about Chopin’s Sonata, for instance, will plausibly have to locate the interoceptive changes quite far into the future. But don’t I often enjoy our favourite music intensely? On the other hand, a relatively faint itch might well be dealt with immediately. The general point is that there seem to be no proportional link between the when and the magnitude of experienced positivity or negativity of an affective experience (Kozuch 2018). A similar problem arises if I take intensity to map onto the magnitude of the predicted interoceptive changes: an IIT proponent would be hard-pressed to argue that the elation of listening to Chopin or watching the sunset is proportional to the magnitude of the expected changes in homeostasis, given that IIT is already over-stretched when it comes to simply linking aesthetic experiences with changes in homeostasis regardless of intensity. In closing, I would like to emphasise that the projects that IIT sets for itself are not undermined by its struggle to provide a mark of the affective. In the end, IIT successfully elucidates the nature of its targets, such as interoception and self-related experiences, and provides a convincing account of the regulative role of affect in the mechanism of instrumental interoceptive inference. The identified shortcoming is in part due to something that might seem like a strength of IIT, which lies in “eliminating any bright line separating emotion (or perception) from cognition.” (S&T, 3). The laudable enterprise of blurring the line between emotion (or more generally, affect), perception and cognition complicates the identification of an aspect that firmly distinguishes the phenomena from each other. However, the intuitive idea that affect, perception and cognition work in close tandem is not per se in conflict with the equally intuitive idea that there are features unique to affective experiences and that distinguish them from perception and cognition. That we cannot straightforwardly extract some of the characteristic features of affective experiences (such as valence) is, in fact, something explicitly acknowledged by S&T: “it remains an open question as to how interoceptive experiences [which IIT equates with affective experiences] map onto the computational machinery of interoceptive inference” (S&T, 3). My aim, then, is not to argue against IIT, but to complement it with the mark of the affective. In the next section, I will look for the mark of the affective in Error Dynamics Theories of affective experience. I will suggest a much-needed amendment to EDT at the end of the section, and in section 5.5, I will develop a revised account that builds on the strengths of both, IIT and EDT, to explain the features of affective experience outlined in section 5.1.
V – ERROR DYNAMICS THEORIES OF AFFECTIVE EXPERIENCE The central tenet of Error Dynamics Theories (EDT) is that positive (negative) valence is equivalent to a positive (negative) rate of Error reduction (a general property of Error dynamics) (Van de Cruys 2017) or, in an analogous free-energy formulation, to a negative (positive) rate of change of free-energy over time (Joffily and Coricelly 2013). In information theory, free-energy is the quantity that bounds the evidence for a model of data (i.e. in the case of organisms, free-energy is greater than the surprise in sensory data). For the scope of this paper, I will discuss things in terms
— 79 —
of Error instead of free-energy, because a fundamental assumption of PP is that free-energy corresponds to the amount of prediction error (Friston 2009, p. 293). According to PP, the central goal of an organism is to minimise Error over time, and Van de Cruys argues that valence plays a key role in the attainment of this goal. This appears intuitive if I consider that affective states usually express our concern and care for things. Now, in PP the main concern of an organism is to reduce Error. The idea is then that affective states are bound up with how I fare in reducing Error. But how? Van de Cruys’ answer is that affective experiences are bound up with Error reduction via a constitutive and unique component of affective experiences: their valence. And this is so because according to Van de Cruys, valence in PP terms just is the rate of Error reduction. From now on I will simply refer to this rate as Rate (short for instantaneous rate of decrease of prediction error, which is the negative of the instantaneous rate of change of prediction error). Rate is, accordingly to EDT, equivalent to valence — positive when valence is positive, and negative when valence is negative. Intuitively, Rate (i.e. valence) reflects the speed or pace with which the organism makes progress in reducing Error. It is this Rate which Van de Cruys proposes to be reflected in the valence of affective experience: If Rate is positive/negative, then affect is positive/negative. Thus, EDT substantiates the intuitive idea that valence is a (nonconceptual) representation of (dis)value to the organism (Levy and Glimcher 2012; Carruthers 2017), an idea dating at least back to Meinong (1917) and at the heart of perceptualist theories of emotions (e.g. Tappolet 2000; Döring 2007). According to EDT, valence is in fact a representation of (dis)value: it represents Rate, the pace at which the agent is making progress in reducing Error over time. While IIT (with its focus on allostasis) seems to emphasise the regulative aspect of affective experience, EDT (with its focus on valence) seems to emphasise their evaluative aspect. An initial worry with EDT’s idea follows from the principle of parsimony: before looking at dynamic variables such as Rate, one should try to use simpler variables to explain valence, such as instantaneous prediction error (i.e. Error). In fact, the initial approach of EDT was to simply consider Error as always “negative in valence” (Van de Cruys and Wagemans 2011). This approach was later abandoned in favour of considering a negative Rate as always negative in valence (Van de Cruys 2017). To justify moving away from his earlier view of equating valence and Error to his more recent view of equating valence and Rate, Van de Cruys builds on the idea that the role of predictive models ensures that PEM leads to homeostasis (based on work in Pezzulo et al. 2015):
Once homeostasis, rather than being reactive, relies on predictive models, errors often do not have direct effect on homeostasis (or fitness). It then becomes equally important to monitor prediction error dynamics, as it is to monitor the errors as such. Mere presence of ºinstantaneous prediction error does not seem to be an adequate basis of emotional valence. Positive affect might still occur for a large instantaneous error as long as this error is (or has been) in the process of being reduced.
-Van de Cruys 2017, 8 Although Van de Cruys does not elaborate further on his reasons for reconceiving valence, it should be noted that a further problem with simply equating Error with valence is that there is always Error involved in any processing, which —if we wanted to simply equate positive amounts of Error with negative valence— would mean that valence is always negative, and that some situations just feel less negative than others. A possible solution to this problem would be to postulate a threshold for (the amount of) Error that is acceptable and that draws the line between negative and positive valence. This move, however, would require positing a further element to the theory, the threshold itself, which would deprive the idea of its parsimony, i.e. of the initial motivation for considering Error simpliciter as a candidate for valence.
— 80 —
Central support for the idea that valence is dependent on error dynamics comes from Joffily and Coricelli’s computational account. In their 2013 paper, Joffily and Coricelli offer a computational model to support EDT, in which they compare two agents in a non-stationary environment — one agent using valence and the other explicitly estimating the volatility of the environment. The agent using valence successfully replicates the behaviour of the other agent and achieves this by representing fewer hidden states and fewer parameters (Joffily and Coricelly 2013). Moreover, EDT’s notion of valence as dependent on error dynamics is in line with Frijda’s view that “pleasure is the positive outcome of constantly monitoring one’s functioning” (Frijda 2007, 82), with the conception of affect as a “neurophysiologic barometer of the individual’s relationship to an environment at a given point in time” (Duncan and Barrett 2007, 1186) and with the control-theoretical approach to emotions of Carver and Scheier, who understand affective experience as an expression of the mismatch between the actual and expected dynamics of a given task (Carver and Scheier 1990; 2001; see also Proust 2015). Finally, EDT also finds support when one interprets various emotion studies in light of PP. For instance, Batson and colleagues argue that a transition from a less valued state (in PP terms, large levels of Error) to a more valued state (low levels of Error) corresponds to positive valence (Batson et al 1992), which is analogous to EDT’s claim that positive valence corresponds to an increase in Rate. Although the support for EDT comes from computational models rather than from neuroscientific studies, it provides a computational story that puts flesh on previous conceptions of valence. In return, EDT finds support from the mentioned work on the evaluative aspect of affective states. To distance itself from IIT, EDT points out that affect by itself is transmodal and distinct from interoception. Addressing IIT, Van de Cruys claims that valence, the basic building block of affective states, doesn’t originate in the inference of the causes of physiological states, but in Error dynamics (i.e. Rate), and because these dynamics are ubiquitous, he claims that feelings “can emerge from any processing, not just that about the body” (Van de Cruys 2017, p. 2). While it might seem at first that this makes IIT and EDT irreconcilable approaches, a closer look at more recent versions of IIT shows this not to be the case. IIT is not committed to the idea that affective experiences can only emerge from the inference of the causes of physiological states. While EDT is a good candidate for accounting for the mark of the affective (with its characterisation of valence), it can very well do so in a way that is compatible with IIT. Furthermore, by providing a computational equivalent of valence, EDT (if correct) might be able to complement IIT. The resulting synthesis (to be developed in chapter 6) provides a plausible account for the elements of affective experience expounded in section 5.2. An interesting question concerns the role that valence plays in the overall goal of PEM. My view is that valence informs the system in order to optimise PEM over time. On the one hand, feelings direct the organism towards homeostasis (a point that is central to IIT proponents, but that Van de Cruys emphasises as well). On the other, feelings direct the organism towards desirable levels of uncertainty (a point that Joffily and Coricelly emphasise). If we grant that evolutionary pressure has made sure that allostasis and PEM are two sides of the same coin (something that IIT and EDT agree on), then valence (Rate) can be used to maintain the policies that minimise Error over time. A key difference between IIT and EDT is that EDT claims that affective experiences play a central role in achieving homeostasis, while IIT claims that affective experiences are homeostasis (more precisely, that they are “the content of the joint set of predictions geared towards allostasis” Seth and Tsakiris 2018, p. 6), with the problematic consequences that we discussed in the previous section. What EDT brings to the table is that Rate also serves the function of optimising levels of uncertainty. Too much uncertainty (e.g. bear scenario) is dangerous for short-term PEM and
— 81 —
requires excessive cognitive resources, too little uncertainty is counterproductive for long-term PEM. The reason to avoid scenarios with too little uncertainty is that they would hamper learning, because the organism wouldn’t be able to appropriately update its models of the world. As Kiverstein and colleagues put it, “being sensitive to error dynamics guarantees that the agent avoids wasting time in places where regularities are either already learned or too complex given the agent’s skill level” (Kiverstein et al 2019, p. 2864). Here the learning rate is an important element. In PP, the learning rate is the pace at which Error changes hypotheses. According to PP, the brain constructs and updates “a vast hierarchy of expectations that overall help regulate the learning rate and thereby optimize perceptual inference for a world that delivers changeable sensory input” (Hohwy 2017, p. 77). Regarding the learning rate, Joffily and Coricelly highlight the following:
An important function of emotional valence turns out to regulate the learning rate of the causes of sensory inputs. When sensations increasingly violate the agent’s expectations, valence is negative and increases the learning rate. Conversely, when sensations increasingly fulfil the agent’s expectations, valence is positive and decreases the learning rate.
-Joffily and Coricelli 2013, p. 1
A necessary addition is that some instances in which the Rate (or valence) decreases over time but stays (slightly) positive (e.g. a highly predictable but changing environment) might mean that a higher Rate is waiting elsewhere and that the present activity should cease (i.e. a low but positive Rate can lead the system to infer that uncertainty will be optimised by chancing activities), and a highly negative Rate might mean that the situation is too complex to learn predictable patterns (e.g. a novice playing the harder levels of Tetris) or, even more important, that the situation is dangerous (e.g. bear scenario) and that the agent should change her behaviour accordingly (not only out of learning rate considerations). This informs the above-mentioned idea that feelings contribute to directing the organism towards optimal levels of uncertainty in order to fulfil PEM over time. This idea is in line with e.g. what Kidd and colleagues have termed the ‘Goldilocks Effect’: “infants actively seek to maintain an intermediate level of information absorption, avoiding allocating cognitive resources to either overly predictable or overly surprising events” (Kidd et al. 2012, p. 6). In PP-terms, maintaining optimal levels of uncertainty is conducive to PEM because it helps the agent to optimise the allocation of cognitive resources, and, as already mentioned, avoid overly high levels of uncertainty. However, there are many situations in which the organism does not seem to avoid low levels of uncertainty, routines and habits being a clear example. If one cleans the house every Saturday it seems unnatural to say that the goal is to optimise uncertainty. Rather, it seems like the goal is to have a clean house over time, even if this goes against the grain of considerations about uncertainty optimisation. Here, the story to follow is not uncertainty optimisation, but the standard PP story, PEM over time. The point is not that uncertainty optimisation substitutes PEM. Rather, PEM subsumes uncertainty optimisation. In other words, uncertainty is optimised to minimise Error over time. The point is that in an EDT formulation, valence is not simply reduced to an engage/disengage signal (of the sort proposed by Prinz 2010). Valence also informs the organism about potential learning opportunities, and as a consequence, the organism might engage in negatively valenced processes if they improve learning in a way that is conducive to PEM. Just as well, the organism will still engage in low-uncertainty processes such as habits and routines, as long as low uncertainty is conducive to PEM over time. Furthermore, conceiving of valence as Rate in the way I have just presented can address the challenge of why there is positive valence associated with activities that do not seem clearly linked to allostasis: sometimes, valence is employed directly at the service of allostasis and sometimes valence is employed at the service of optimising uncertainty.
— 82 —
Nevertheless, there is a shortcoming in the version of EDT defended by Van de Cruys: it does not account for a central element of affective experiences, namely, that feelings are conscious phenomenal states. Van de Cruys does not simply think that valence determines affective experiences, but also that, since valence is usually understood as a phenomenal aspect of affective experience, changes in Error are consciously experienced as valence. A problem with this is that affective feelings are conscious and, according to the standard view of PP, Error is not part of conscious experience. By extension, changes in Error qua Error cannot be part of conscious experience. Within the PP framework, the content of conscious experience are predictions. In Hohwy’s words, “Conscious perception is determined by the hypotheses about the world that best predicts input and thereby gets the highest posterior probability. More specifically, since the inversion of the generative model is implicit, what is conscious is the interconnected set of currently best performing predictions down throughout the perceptual hierarchy” (Hohwy 2012, p. 201). Of course, Error is constantly used to update predictions but in the end what subjects consciously experience is not Error but predictions. To defend the idea that predictions determine the content of conscious experience, Hohwy follows an abductive argument. Within the PP framework, the conception of conscious content as predictions is the best explanation for a series of explananda arising from empirical studies in cognitive psychology, psychophysics and psychopathology, as well as from a conceptual analysis of our folk understanding of conscious experience. In particular, Hohwy argues that the concept of conscious content tracks characteristics of conscious perception such as “binding, penetrability, reality testing, illusions, inextricably rich experience, and first person perspective” (Hohwy 2013, p. 202). A good example of the concept of conscious content as predictions at work is provided by Hohwy’s explanation of binocular rivalry (Hohwy et al. 2008). In the most common version of binocular rivalry, one eye is presented with a picture of a house and the other eye is presented with the picture of a face, and conscious experience switches between the two percepts. Hohwy’s PP explanation is that the likelihood of the combined face-house blend hypothesis cannot overcome “the exceedingly low probability that a face and a house could co-exist in the same spatiotemporal location … so the hypothesis that is selected, and which determines perception, is either the face or the house hypothesis” (p. 21-22, Howhy 2014). That predictions constitute the content of conscious experience is the standard view in PP. For instance, Barret writes that “once prediction error is minimised, a prediction becomes a perception or an experience” (Barrett 2017, p. 12). This view extends beyond exteroception. Hohwy argues that this view also provides a good explanation for the role of expectations for bodily sensation (Hohwy 2011, p. 271), namely, that innocuous stimuli that are expected to be painful tend to be experienced as painful (Brown et al. 2008). Importantly, IIT (in contrast to EDT) is very clear in upholding the view that predictions determine the content of conscious experience: “An important challenge in this context is to identify which aspects of inference support specifically conscious emotional experience, with predictions (rather than prediction errors) being the preferred vehicle” (Seth and Friston 2016, p.5). This view poses a challenge for EDT. The resulting argument goes like this: P1) Only predictions constitute the content of conscious experience P2) Valence is a content of conscious experience P3) Rate is not a prediction C) Rate is not equivalent to valence A viable solution to this issue is to extend EDT by casting valence in terms of predictions. Here, I propose that a given set of predictions can include predictions of Error dynamics. That is, the predictions that the subject experiences can include predictions of the (expected) Rate. Therefore, if,
— 83 —
as EDT claims, feelings co-vary with properties of Error dynamics, what I experience as these feelings cannot come directly in the form of Error dynamics but must come in the form of the system’s predictions about Error dynamics. The resulting view is that the valence of affective experiences is determined by the prediction of the Rate, Expected Rate from now on. This means that positive (negative) phenomenal valence corresponds to the Expected Rate being positive (negative). On the other hand, (instantaneous) Rate might be best understood as unconscious valence (e.g. Berridge and Kringelbach 2015). It might seem that adding “another layer” of computations (i.e. from Rate to Expected Rate) might bite the bullet at the expense of parsimony. However, the opposite turns out to be the case. Taking a formal account of volatility and learning rate under Bayesian inference as a starting point, Perrykkad and Hohwy argue that the rate at which a prediction error is minimised is partially determined by the precision ascribed to that prediction, so that if an organism holds expectations about precision (a central tenet of PP) it then holds implicit expectations about Rate:
Technically, the rate at which a prediction error is minimised is partially determined by the precision (or confidence) ascribed to that prediction. Intuitively, we update our beliefs faster and more dramatically when we believe our data to be more reliable and when we suspect the world is frequently liable to change […] This means that those creatures whose internal models allow them to make predictions about this precision implicitly hold (sub-personal) beliefs about the expected rate of error-minimisation, conditioned on the data they choose to sample. Given the inverse relationship between precision and ambiguity, we can associate beliefs about the ambiguity expected under a given policy with beliefs.
-p. 3, Perrykkad and Hohwy 2020
Perhaps some creatures hold explicit beliefs about Expected Rate, but as long as the organism is calculating expected precision, it is, by extension, implicitly estimating Expected Rate — making Expected Rate more parsimonious than Rate simpliciter. This elaboration of EDT makes Expected Rate the best candidate for valence and, more importantly, for the mark of the affective, which is a necessary element to account for affective experiences.
VI – CONCLUSION I started this chapter by outlining the characteristics of affective experiences which a PP theory should aim to account for. Affective experiences are phenomenally conscious. A central aspect of this phenomenal nature is phenomenal valence, which I argued is the most promising candidate for the mark of the affective. Another dimension of affective experiences is their level of arousal. Furthermore, affective experiences motivate and guide behaviour and are intentional, that is they are about something. Traditionally, their intentionality is thought to comprise a specific object they are directed at, say a body part or a bear, and a feeling-specific property assigned to this object by the affective experience. These are respectively called the particular and formal object of the feeling. Secondly, I reviewed the existing theories of affective experience in the PP framework and divided them into two families: IIT and EDT. I showed the virtues and shortcomings of the two families of theories in accounting for some of the outlined features. Regarding IIT, I argued that while it provides an excellent account of the role of affect in instrumental interoceptive inference and allostasis, it does not provide an adequate mark of the affective. EDT, on the other hand, offers promising computational accounts of valence, and by extension of the mark of the affective. However, EDT disregards a basic premise in PP: that it is predictions, and not error, that determine
— 84 —
conscious experience. If we want to account for feelings as phenomenally conscious affective states, we need a theory that casts feelings as predictions, and not errors. The next chapter sets itself the task of providing such a theory.
— 85 —
6 . T H E A F F E C T I V E I N F E R E N C E T H E O R Y O F
F E E L I N G S
I – INTRODUCTION The previous chapter concluded that Expected Rate (short for expected instantaneous rate of decrease of prediction error) is the best candidate for valence within the Predictive Processing (PP) framework and, by extension, the best candidate for the mark of the affective. This chapter will build around the notion of Expected Rate introduced in the previous chapter to propose a revised PP-theory of affective experiences, the Affective Inference Theory (AIT), which cannot only accommodate the evaluative but also the motivational and intentional character of feelings. In what follows, I will show how AIT can account for the commonalities of affective experiences outlined in chapter 5:
• Affective experiences are phenomenally conscious • Affective experiences are valenced, • Affective experiences vary in their level of arousal, • Affective experiences are intentional states with a particular and a formal object, • Affective experiences motivate behaviour.
Section 6.2 equates Expected Rate with phenomenal valence and felt arousal with the expected changes in Rate. Section 6.3 elaborates on the intentionality of feelings by building on the notion of regularity within PP: I argue that the particular and formal object of a feeling emerge as the inferred hidden causes of variation in predictive success. In section 6.4, I extend the emerging account to explain how affective states motivate and guide behaviour, building on previous PP work on action policy selection. Section 6.5 recapitulates the central aspects of the account and applies it to metacognitive feelings.
II – PHENOMENAL VALENCE AND FELT AROUSAL AIT’s central tenet is that the valence of affective experiences corresponds to Expected Rate, so that positive (negative) feelings correspond to Expected Rate being positive (negative). Based on this conception of valence, AIT integrates insights of Interoceptive Inference Theories (IIT) and Error Dynamics Theories (EDT) to develop an account that is compatible with the core of both theories while trying to rise to their challenges. Importantly, Expected Rate (unlike EDT’s Rate) is a plausible candidate for phenomenal valence, and consequently, it provides a plausible mark of the affective in computational terms (unlike IIT’s interoceptive inference).
— 86 —
Equating Expected Rate with valence explains the way in which affective experiences are valenced, and also the way in which affective experiences are conscious: only predictions constitute the content of conscious experience, and because Expected Rate is a prediction, it can populate consciousness in the form of phenomenal valence, the core of affective experience. By recasting valence in terms of predictions, AIT sets itself apart from EDT. As for IIT, the notion of Expected Rate can be seen as the computational ingredient lacking in the machinery of instrumental interoceptive inference and thus complimentary with its theoretical approach (Seth and Tsakiris 2018). AIT is compatible with the idea that instrumental interoceptive inference plays an essential role in experiences of embodied selfhood, such as “experiences of being an embodied organism, experiences of mood and emotion, pre-reflective experiences of selfhood and ‘mineness’, explicit self-awareness, metacognitive insight, reflective self-awareness, and social aspects of selfhood” (Seth and Tsakiris 2018, p. 10). To define valence in computational terms that are orthogonal to interoception is not to deny the deep connection between interoception and affective experience. Somatic changes are essential queues for predicting changes in Rate, and Expected Rate plays an essential role in maintaining homeostasis, a point that AIT, EDT and IIT all agree on. Finally, as will become clear in section 6.3, AIT defends that affective experiences are transmodal experiences with an important somatic component. A good example of AIT at work are affective experiences like those of listening to music introduced in the third section of the previous chapter. A major current in aesthetics holds that affective experiences when listening to music arise from the composer’s manipulation of expectations (Meyer 1956, for empirical support see Huron 2006). AIT can provide a computational PP account of affective experience in line with the view of music as expectation manipulation. Music has constantly evolving patterns that result in unique error dynamics. Say I have finally “gotten used” to a pattern, the music changes, and the organism predicts negative Expected Rate (negative valence); this negative valence then continues until the organism becomes more confident in its capabilities to predict the new pattern, resulting in a positive Expected Rate (positive valence), until the music changes again, and so on and so forth. Importantly, AIT offers a model for how Expected Rate can operate at several timescales at once. If the organism expects to be about to generate a good model of the underlying structure of a piece of music (phenomenologically, the anticipation of insight), this will result in positive Expected Rate (i.e. positive valence directed at an abstract, longer term level) even when the Expected Rate of the immediately incoming auditory stimuli is negative (i.e. negative valence directed at a concrete, short-term level). This tension between short-term and long-term Expected Rate underlies much of the phenomenology of experience of music, which is laden with emotional tension, fulfilment, disruption and anticipation.22 We can now explain two of the characteristics of feelings that I argued a theory of affective experience needs to account for: that feelings are conscious, and that feelings have valence. Arousal, another characteristic of feelings, can be approached in an analogous way to valence. In Joffily and Coricelli’s model, the second time-derivative of valence has explanatory power regarding the dynamics of basic forms of emotion (Jofilly & Coricelli 2013). Based on agential simulations, they argue that valence and its first derivative constitute the two central dimensions along which feelings vary, so that all feelings can be plotted on a two-dimensional graph having valence and its derivative as axes. If we follow the widely supported view that the two essential 22 This is in line with the early EDT idea that “prediction errors at the level of style (perceptual ones) sometimes can be
resolved on the level of meaning” (Van de Cruys and Wagemans 2011, p. 1053). However, Van de Cruys and Wagemans fail to flesh out this idea, claiming at once that “a central assumption of our theory is that prediction errors are always to some extent emotional, more specifically negative in valence” (p. 1047) and that “this playful and safe as-if context of art, where our guards can be lowered and our actions suspended, allows for the usually negative prediction errors to be enjoyed” (p. 1041), without explaining how negative prediction error suddenly turns into positive phenomenal valence in the context of aesthetic experience.
— 87 —
dimensions of affective experience are valence and arousal, it seems that if the Expected Rate corresponds to valence (as was argued in the previous chapter), then, by analogy, the best candidate for arousal within PP is its derivative23. Therefore, in AIT, arousal corresponds to the expected value of the derivative of Rate, or, put simply, Expected Change in Rate. Intuitively, Expected Change in Rate denotes how rapidly Rate is expected to change. If Expected Change in Rate is high, then Error is expected to change fast, and vice-versa. Arousal is then a direct expression of the “acceleration” in the agent’s performance at reducing error. If arousal is high, then the feeling would be experienced as more “fast-paced”, sudden or exciting. Moreover, this conception also makes sense in terms of the link between arousal and resource allocation, and between arousal and action. The faster (slower) Error is expected to change, the more (less) resources are required to deal with this potential change (for the link between arousal and resource allocation, see Hinojosa 2009). And the faster (slower) Error is expected to change, the faster (slower) and more (less) intense the action response will tend to be (for the link between arousal and action, see Blakemore and Vuilleumier 2017). In AIT, valence (Expected Rate) and arousal (Expected Change in Rate) are seen as the two constitutive dimensions of feelings, echoing Russell and Barrett’s work (Russell and Barrett 1999; Russell 2003). Felt valence and arousal emerge then as nonconceptual proxies for value where value is an agent’s expected success in predicting the world (Expected Rate), and the expected stability of this success (Expected Change in Rate).
III – INTENTIONALITY With AIT, we can now explain three of the characteristics of feelings outlined in section 2 of the previous chapter: that feelings are conscious, that feelings have valence and that feelings vary in their level of arousal. Now, we turn to another aspect of affective experiences: intentionality. As I argued in chapter 5, phenomenal valence is the best candidate for the mark of the affective, partly because it grounds the other features of affective experiences. We can build on AIT’s notion of phenomenal valence to provide an account of the intentional and motivational dimension of affective experiences. First, let us focus on intentionality. Felt valence emerges as a nonconceptual representation of value fleshed out in terms of Expected Rate, an agent’s expected success in predicting the world (Carruthers 2017; Van de Cruys 2017). When I say that valence can be computationally captured as Expected Rate, it is important for the system to also know what (and in what way) is causing (or modulating) the changes in Expected Rate. In other words, it is important that the feeling in question captures what is most likely causing the changes (which I equate with its particular object) and the most likely reason why it is causing these changes (which I equate with its formal object). Remember that a feeling represents its particular object as having a feeling-specific property, i.e. its formal object. This is so because representing the connection between the valence and its modulator is crucial for the system to effectively deal with the sources of perturbation in predictive success. One way in which a predictive system can naturally establish these connections is by tracking and predicting regularities, resulting in an experienced connection of the features exhibiting the regularity. Wiese’s recent regularity account of PP illuminates this notion: “When the brain tracks a regularity that is 23 The correspondence between the second derivative and arousal is more tenuous than the correspondence between the
first derivative and valence. While this correspondence is supported through Joffily and Coricelli’s computational work, this conception needs to be cemented through further research. This conception of arousal should be put in competition with alternative conceptions (e.g. expected variance of Rate) in further computational models, and eventually supported through direct empirical work by inferring what psychological hypotheses one should expect out of this conception and testing those hypotheses.
— 88 —
predictive of different features (or of different objects or events), there will be an experienced connection between those features (or the respective objects or events). I can then say that the regularity connects those features (or objects or events)” (Wiese 2018, p. 259). The idea behind the regularity account is that some features are strongly predictive of other features based on a robust observed correlation between these two sets of features. A good example of regularity is binding. Binding is a term usually employed to refer to the binding of multiple intra- and cross-modal features to a single object, such as the binding of the visual information about the shape and the colour of a bouncing ball (intra-modal) or the binding of the ball’s visual features and the auditory features of it bouncing (cross-modal) (see e.g. Hommel 2004). Now, a PP system does not have to first process stimuli separately and then bind them. Instead, it uses causal inference embodied in the parameters of its generative model to assume bound attributes and then “predicts them down through the cortical hierarchy. If they are actually bound in the states of the world, then this will minimise prediction error, and they will be experienced as such” (Hohwy 2012, p. 6). In short, binding in PP terms is simply the exploitation of statistical regularities in the form of Bayesian priors to predict a strong connection between two or more variables. In the case of the intramodal binding of the shape and colour of a bouncing ball, these two features are predicted to have a strong, localised connection, and this regularity results in a high degree of experienced localised integration. The shape and colour of the ball are experienced as the same object, localised precisely in space. In the case of the cross-modal binding, the visual and auditory features of the bouncing ball are related, but not as strongly as its shape and colour, and although both stem from the ball, they occupy different regions in space. This results in the visual and auditory features being experienced in synchrony and as related to the same object, so that this cross-modal aspect of experience has a high degree of integration, but a lower one than the intra-modal aspect of the experience. Thus, if there is an observed regularity (e.g. causality or correlation) between an object and changes in Expected Rate, this object will be experienced as the intentional object of the feeling with a certain feeling-specific property that explains these changes. Based on a tracked regularity, the particular object of a feeling emerges as the inferred hidden causes of a variation in predictive success. What follows is that formal and particular objects, valence and arousal are all integrated into the phenomenal whole of a specific feeling. When it comes to feelings, the connection between their objects and their affective phenomenal dimensions exhibits statistical regularities that are useful for prediction, but weaker than the statistical regularities underlying sensory binding. This is because the features between which these regularities are observed are of an abstract, transmodal nature, involving, for instance, multi-modal perception of objects (which can be whole situations) together with expected interoceptive consequences and with whatever somatic cues support the inferential process. This results in an affective experience that is often concurrently coded in many modalities (exteroception, interoception) and, at the same time, has a substantial degree of integration, but a lower one than typical experiences of sensory binding (cf. Frijda 2005). The particular object of a feeling is what is predicted to be the most likely cause of the expected Error dynamics. In the example of a person walking through the forest and becoming afraid when spotting a bear, it is the bear that is predicted to cause a change in Rate and it is therefore the bear that is experienced as the particular object of the ensuing feeling, i.e. the object towards which the feeling is directed. In turn, the formal object, i.e. danger, corresponds to the Expected-Rate-relevant properties of said particular object, bound up with other (perceptually) predicted properties such as the size of the bear. In this context, a formal object works as a predictive model of the expected changes in Error dynamics caused by a particular object. The system has priors about the presence of certain relevant properties given such and such input (e.g. given the valence, particular object and context of a situation).
— 89 —
Given a certain input, the system predicts the presence of a formal object (as the property of a particular object), similarly to how the system predicts the presence of the colour red (as the sensory properties of a certain object), given specific events on sensory sheets. This analogy raises the following question: are the properties that are formal objects of feelings different from sensory properties that bottom out in sensory sheets? Some formal objects such as bodily damage or comfort in the case of pain or affective touch respectively seem to be very similar in that they have dedicated sensory pathways (Auvray et al. 2010; McGlone et al. 2014). Other formal objects such as danger or funniness are different in that they are more abstract in a specific way; we do not have dedicated sensory receptors upon which (precursors of) such properties could impinge. In any case, formal objects are inferred high-level properties that is some cases bottom out in sensory sheets (e.g. bodily damage) and in others don’t (e.g. funniness). Going back to the example of the bear encounter, the bear is expected to be dangerous, and danger is the formal object of this experience. Unlike the colour of the bear’s fur, danger is not part of the sensory content of the experience. Rather, danger is a property that is affectively framing the experience itself. Due to the ability of a PP system to establish experienced connections based on tracked regularities, this inference-based intentionality will be phenomenal in nature. The emerging picture allows for a natural way to understand moods such as depression, anxiety or elatedness which are seemingly objectless or, alternatively, directed at the whole world. Note that felt moods are peculiar when it comes to their particular objects while apparently having formal objects similar to directed emotions such as sadness, fear or joy (cf. Mendelovici 2014). Against the background of AIT, moods can be understood as predictions of error dynamics whose specific cause and, thus, particular object cannot be inferred. This can happen for a variety of reasons. The system might simply fail to inferentially identify a particular cause. This may be due to the inaccessibility of the modulators of Rate when they, for instance, come in the form of unobservable neurochemical fluctuations. Sometimes, however, there is simply nothing in particular that is responsible for a change in Rate. Instead, there might be a whole series of events (and not one in particular) where the subject performed well or poorly at reducing Error, leading accordingly to a change in Expected Rate. In such cases, the mood can be thought to directly (and not through a particular object) represent how the subject expects to fare at reducing Error in the world. As a consequence, the subject will experience the world as generally deprived of opportunities (to reduce Error), full of threats to opportunities (to reduce Error) or as full of opportunities (to reduce Error) (cf. de Haan et al. 2013). In other words, moods are affective experiences that do not lack intentionality altogether but parts of the affective intentionality (i.e. the inferred cause of expected changes in Error dynamics is not clearly present in the phenomenology in the form of a particular object). This seems to us as a natural conceptualization of depression, anxiety or elatedness. Similar things could be said for other feelings that appear relatively undirected, such as feeling tired, relaxed or lascivious. So far, we have an emergent theory in which the PP system generates predictions about Rate (valence) and Changes in Rate (arousal), determining a feeling that, through the prediction of regularities, gets bound to a particular object and a formal object (the inferred cause of expected changes in Rate) into a global phenomenal whole. Because this process happens in a feedback loop, feelings get constantly updated. The system takes into account input from different modalities and at different levels of the predictive hierarchy to modify its predictions, thus making feelings context-sensitive (Proust 2015).
IV – MOTIVATION
— 90 —
A final elaboration in order to account for the target explananda is the connection between feelings and behaviour (or action — I use the terms interchangeably, reflecting the usage in the PP literature). As I mentioned earlier, in the PP framework, action happens to fulfil proprioceptive predictions. My claim here is that feelings bias action selection, so that specific behaviours become more likely. This biasing happens at different levels and timescales. It is likely that not only immediate actions will become more probable because of feelings but also certain actions in the future or distal predictions such as goals (see Pezzulo and Cisek 2016). In PP we can think of goals as high-level predictions that influence action in the following way: “I imagine a future goal state as actual, then use Bayesian inference to find the set of intermediate states (which can now themselves be whole actions) that get us there” (Clark 2015, p. 120). It is also important to emphasise that the dynamics involved here are non-linear. Because all of this happens in a feedback loop, feelings will modify the likelihood of actions, but actions will also modify feelings in turn, by modifying the error dynamics and with them, the predictions (e.g. valence) determining the feelings. In general, engaging in actions spans predictive trajectories in whose course feelings can arise in a regulative fashion (cf. Proust 2013). What this means is that in performing the component steps that implement an action, error dynamics unfold, leading to possible alterations in Expected Rate, giving, in turn, rise to regulative feelings in the form of predictions about error dynamics (i.e. Expected Rate). This conception of the influence of feelings on behaviour resonates with the concept of action tendencies, the idea that a given emotion will make a set of actions more likely than others (Frijda 1986). To fully address the complexity of the feeling-behaviour relation, we need to elaborate on the notion of formal objects introduced in the previous section. We equated formal objects with the hypothesised Rate-relevant properties of the particular object. These hypothesised properties (e.g. the danger of a bear in the case of fear) work as a partial model of the particular object that can help the agent direct her actions towards the desirable values of Rate (i.e. the values that are expected to be conducive to PEM over time). Formal objects should therefore not be viewed simply as passive descriptions of the particular object, but also as calls for action in the form of action tendencies. The property “danger” is not simply the likelihood of the bear deciding to attack, but also the affordance to escape, or to raise one’s arms trying to look big (cf. Proust 2015). The prediction of the property “danger” is thus a “pushmi-pullyu representation”, both descriptive and directive (Millikan 1995). This makes sense when we see that formal objects are high-level hypotheses (i.e. a joint set of predictions), and that in PP, high-level hypotheses are “intrinsically affordance-laden: they represent both how the world is and how I might act in that very world” (Clark 2015, p. 187). Given a particular object, a context, and a certain valence, the formal object is a way of using prior knowledge to predict how a situation will unfold (descriptive properties) and to regulate behaviour (directive properties) in order to direct uncertainty to the desired levels. The notion that formal objects are not only descriptive but also directive, links with the work developed in PP around action policies. Perrykkad and Hohwy give a good definition of action policies: “A policy is a set of possible actions (or individual control states) that have been grouped together by the individual for its history of success as a strategy to reduce prediction error when faced with situations with learned commonalities, which cue success for that policy.” In the same breath, they also specify that expectations for “rate of prediction error minimisation will impact on policy selection” (Perrykkad and Hohwy 2020, p. 3). Accordingly, based on previous dynamic patterns of Rate, formal objects arise to predict how Rate is going to unfold (descriptive) and (based on expected Error dynamics) to regulate actions to minimise Error over time (directive). Here, we are just extending to feelings the PP idea that “agents are assumed to infer ‘what is the current state of the world’ and ‘what are the best actions to pursue’ based on the same generative model of the environment” (Schwartenbeck et al 2018, p. 3). According to AIT, when it comes to inferring the dynamics of Error (i.e. Expected Rate) and what are the best actions to pursue (based on Expected
— 91 —
Rate), then we are talking about formal objects that are at once affective models and affective action policies. Here, it is worth revisiting some of the discussion about IIT from the previous chapter, where I noted that, for IIT, the central function of instrumental interoceptive inference is the model-based regulation of interoceptive variables. According to AIT, this function is fulfilled through the modelling of Expected Rate (the evaluative aspect of affective experience), which, as we have just seen, involves action policies (the regulative aspect of affective experience). This is possible because, as we saw in the previous chapter, the organism employs Expected Rate not only to optimise uncertainty, but also to direct the organism towards homeostasis. Importantly, the regulation does not only happen through proprioceptive actions but also through intero-actions (i.e. autonomic reflex arcs). Pezzulo et al. provide a useful account of motivation and control, based on which we can further specify the path from feelings to behaviour via formal objects (and precision). They argue that control (goal selection) and motivation (goal prioritization) are two intertwined parts of the same process of active inference. So much so, that they prefer to talk in terms of controlled motivation, or motivated control. A deep hierarchy generates goals (action policies, in the terms of our present discussion), which influence the value of different actions (i.e. motivation), but the maintenance and propagation of these goals are dependent on expected precision, which is again dependent on motivation (Pezzulo et al. 2018). As for formal objects qua action policies, this is consistent with the idea that formal objects work as a partial model of the particular object that can help the agent direct her actions towards the desirable values of Rate — a given action policy (formal object) increases the expected precision of the actions that are inferred to fulfil said action policy, making those actions more likely. What is interesting in Pezzulo et al.’s account is that it shows how Expected Rate can then influence the formal object and the resulting action in a dynamic way. In their own words, “this view may help understand the multifarious phenomenology of goal processing, such as the positive emotions associated with progress towards the goal (anticipation, enthusiasm) and the negative emotions associated with failures (disappointment, regret), in terms of increased (or decreased) confidence that the selected policy will achieve the desired goals” (Pezzulo et al. 2018, p. 304). Think now of formal objects qua models: If there are unexpected changes in Rate, this will decrease the precision of the formal object as a model of Rate, so that either the model will need to adapt (again, in an evaluative-regulative fashion, as with increasing frustration about a difficult task) or it will be substituted by a different model (e.g. going from anticipation to disappointment). Going back to the idea of affordance competition introduced in chapter 5, we can think of competing potential actions forming an ‘affordance landscape’ (see Pezzulo and Cisek 2016 for a discussion of this ‘affordance landscape’ hypothesis and for a mapping of its ‘hierarchical affordance competition’ process to brain structure). When feelings emerge, they transform said landscape (qua affective models) and (qua action policies) guide the organism in navigating through it. In a recent paper, Kiverstein and colleagues suggest that valence lies directly in the affordances of the environment, so that a nearby apple will have positive valence if it looks tasty and negative valence if it looks rotten (Kiverstein et al 2019). In the present view, valence lies in expected error dynamics, namely Expected Rate, and only gets tied to environmental objects (i.e. particular objects) and to affordances (i.e. the directive aspect of formal objects) derivatively through the prediction of regularities. AIT’s regularity account, together with the conception of formal objects as both descriptive and directive models, shows how affordances and environmental objects can gain their apparent valence, which is better understood as object valence derived from phenomenal
— 92 —
valence. In consonance with a recent computational model of valence and policy selection, it is better to think of affordances as “affectively charged” (Hesp et al. 2019).24 The idea that affect is linked to affordances is particularly compelling when we consider cases of prototypical negatively valenced feelings, as in the example of encountering a bear. In said example, the formal object “danger” is not simply a description of the bear, but a display of affordances that are experienced as directive pushes for action, such as raising one’s arms, or turning around and running. However, it does seem that not all feelings are clearly linked up to specific affordances in this way. Take for instance cases of positive affective experiences: What is the affordance of pleasant relaxation or simple joy or happiness? Some formal objects are best understood as affordance-laden hypotheses while others, perhaps, are not. Here the PP framework offers a natural way to understand the gradation between descriptive and directive predictions. It is useful to think of the formal objects of feelings associated with a lot of action-relevant content as bona fide action policies (e.g. fear), but there is a gradation to the amount of action-relevant content a prediction can have. In other words, some feelings qua formal objects are more directive than others. Additionally, we need to keep in mind that some affective experiences will be immediately and obviously directive (e.g. running away) while others will be immediate but much more subtle (e.g. recalling pleasant memories). Yet others will only raise the probabilities of situationally removed future behaviours (e.g. seeking out more works by a given artist in the future).
V – METACOGNITIVE FEELINGS
Through all he said, even through his appalling sentimentality, I was reminded of something — an elusive rhythm, a fragment of lost words, that I heard somewhere a long time ago. For a moment a phrase tried to take shape in my mouth and my lips parted like a dumb man's, as though there was more struggling upon them than a wisp of startled air. But they made no sound and what I had almost remembered was uncommunicable forever.
— Scott Fitzgerald. The Great Gatsby The feeling described in the above excerpt is an incidence of the tip-of-the-tongue experience (ToT), a feeling that occurs when a person is trying to retrieve an item that is temporarily inaccessible (Brown and McNeill 1966, Schwartz and Metcalfe, 2011). When most people think about feelings, things like fear, anger or sadness might come to mind, not necessarily the ToT experience. ToT seems to belong to a particular category that has been the topic of much recent discussion in psychology and philosophy: Metacognitive feelings (Koriat and Levy-Sadot, 1999; Arango-Muñoz, 2013), sometimes also referred to as cognitive (Greifeneder, Bless and Tuan Pham, 2010), epistemic (Proust, 2007, 2008; de Sousa, 2008) or noetic feelings (Dokic, 2012). Metacognitive feelings have been characterized as “feelings concerning the subject’s own mental capacities and mental processes” (p. 97, Arango-Muñoz and Michalean, 2014), as “feelings that enter into the epistemic processes of inquiry, knowledge and metacognition” (de Sousa, 2008), or as “feelings about knowing” (Clore, 1992). A big open question about metacognitive feelings concerns their underlying mechanism. We know that metacognitive feelings correlate with certain process properties, most notably process fluency (Alter and Oppenheimer 2009, Whittlesea and
24 In fact, the computational model of Hesp and colleagues can be seen as an extension of the principles behind EDT
(Joffily and Coricelly 2013) to the inference of future states of subjective fit (i.e. affective inference) and consequent action selection. Whereas my work is mainly philosophical in nature, their work is mainly computational in nature, but I see this convergence as an auspicious sign.
— 93 —
Williams 1998, 2001a, 2001b), but there is no encompassing theory of how they emerge. We know that metacognitive feelings influence behaviour, but there is no definite notion of the way in which they do so. In this section, the AIT will shed light on how metacognitive feelings emerge and how they affect behaviour. To get a better picture of what metacognitive feelings are, let us start by exploring some feelings that are widely regarded as metacognitive, namely, the feeling of knowing (FoK), the ToT experience and the feeling of confidence (FoC). For the FoK, consider the British TV programme University Challenge, in which teams from two different colleges compete to respond to a series of trivia questions. The challenge is double: to answer correctly, and to do so before the other team. If a participant feels that she knows the answer, she has to push the buzzer faster than the other team, with participants often pushing the buzzer before the question itself is finished. With questions as intricate as “in family relationships, what four-word term is used to describe a child of the child of a first cousin of either of one’s parents in relation to oneself?” a participant often has to rely on her “gut feeling” to decide whether to push the buzzer or not, since if she waits until she is sure that she knows the answer, the other team will push the buzzer before her. The “gut feeling” in question is the FoK. Upon hearing the question, a feeling of knowing the answer emerges often before the answer itself, prompting the participant to push the buzzer. A related feeling is the ToT experience. This exasperating experience happens when you feel you know something (e.g. the name of the actor playing Ron Wesley in the Harry Potter films), but you are unable to recall the information in question (“Rudolph Green?…”), and you have the persistent feeling that you are about to retrieve the information (“…not that. Rupert Green? It’s something like that. Robert Green?”) but fail to do so. In other words, you feel that said information is “on the tip of your tongue”. As for the FoC, which I already mentioned in chapter 4, this is the kind of feeling that guides people taking a multiple-choice test in which wrong answers are penalised with negative points. The examinee might think that answer “a” is more likely than the other three possibilities, but whether she chooses to tick it or to leave the question blank will be influenced by how sure she is of the answer. If she feels confident that “a” is the right answer, then she will be more likely to choose to tick it instead of leaving the question blank. The feeling guiding her behaviour here is the FoC. In what follows, I will outline some important characteristics of metacognitive feelings. To begin with, a central characteristic of metacognitive feelings is that —although they often reside in the fringe of consciousness (James 1890, Mangan 1993, 2000, 2001; Reber et al 2002: Norman et al. 2010) and may sometimes be considered “background feelings” (Colombetti 2011, 2014)— they are conscious states (Koriat 2000; Koriat and Levy-Sadot 2000). Furthermore, metacognitive feelings concern the subject’s own mental or cognitive capacities and processes. This is essentially why they are called metacognitive feelings. For instance, the content of the FoK could be re-described as “I know this”, i.e. as informing the subject about her ability to undertake a given mental task such as memory retrieval (Dokic, 2012). Another important aspect of metacognitive feelings is that they are conscious affective states. On the one hand this is to say that their phenomenology comprises a positive or negative valence (Topolinski and Strack 2009; Proust, 2015; Carruthers 2017), i.e. they feel positive (e.g. FoK) or negative (e.g. ToT). This component can be seen in part as an evaluation of the quality of the mental capacity or process with which the metacognitive feeling in question is concerned. On the other hand, their phenomenology is also characterized by felt arousal, i.e. during a metacognitive feeling the subject feels a more or less localized decrease or increase in her level of activation, energy or excitement (Morris 2008; Proust 2015; Colombetti 2018).
— 94 —
Of course, phenomenal valence and felt arousal are characteristics that metacognitive feelings share with all affective states. As for the particular function of metacognitive feelings, Arango-Muñoz and Michalean make the following suggestion: a subject lacks direct access to and thus cannot know for certain her own mental capacities and processes. The role of metacognitive feelings might be to enable the subject to deal with this uncertainty of the mind (Arango-Muñoz and Michalean, 2014). This makes sense specially in the light of the AIT notion that feelings direct the organism towards desirable levels of uncertainty. A further important characteristic of metacognitive feelings is that they motivate and influence the subject’s bodily and, importantly, mental behaviour and actions in characteristic ways (Proust 2008, Norman et al. 2010). The FoK might prompt a given participant of University Challenge to press the buzzer, and the feeling of forgetting (FoF) might induce a subject to try to remember whatever it is she might be forgetting, by, for example, mentally going through all of the objects she needs to pack before leaving the house (Halamish et al. 2011, Arango-Muñoz 2013). These considerations concerning the involvement of metacognitive feelings in the management of cognitive uncertainty and (mental) behaviour reflect well a common theme within the literature: metacognitive feelings are considered part of meta-cognitive monitoring and control, i.e. they stem from mechanisms that monitor and control cognition (see Proust 2014 for a review). In order to sharpen the characterization of metacognitive feelings, it is useful to go back to the concepts of particular and formal objects. A formal object is the property which a feeling represents an object as having, but what property a metacognitive feeling represents depends on whether we are considering the personal or the sub-personal level (Dennett, 1969). For example, at the personal level, the property of the FoK is “something being known,” but at a sub-personal level, we might need to look for properties such as processing fluency (Alter and Oppenheimer 2009). Similarly, the particular object of the FoK at the personal level might be a given mental, possibly propositional content (e.g. the name of the capital of Nepal), while at the sub-personal level it might the processes implementing memory retrieval or the ability to engage in such a task. In general, the formal objects of metacognitive feelings will be, on the personal level, cognitive properties such as confidence, rightness, being known or forgotten, and on the subpersonal level, vehicle-properties (i.e. process-properties, like fluency). And the particular objects of metacognitive feelings will be (on the sub personal level) cognitive processes, abilities and dispositions, which on the personal level will often, but not always, appear to be (the content of) cognitive states such as thoughts and beliefs. Metacognitive feelings will tend to have both cognitive processes and cognitive properties as particular and formal objects, but there is no need to offer a restrictive definition of metacognitive feelings following this characterization. The reason for this is that while sketching a characterization is useful to clarify our explanatory target, it is far from clear whether or not metacognitive feelings form a natural kind the properties of which can serve to give a fitting definition of this class of affective experiences (see Meylan 2014 for a discussion). As for individual metacognitive feelings, they certainly tend to be specific to the cognitive activity or process that the mechanisms that give rise to the metacognitive feeling in question are monitoring (e.g. memory retrieval). Which activity or process that is will probably play a substantial role in determining not only the particular but also the formal object of the metacognitive feeling (e.g. the property of something being known), but again, there is no need for the purposes of this chapter to argue for or against the individuation of different metacognitive feelings in terms of their formal objects. Overall, we can see that, as one would expect, metacognitive feelings share all of the essential features of affective experiences. That is, they exhibit phenomenal valence and felt arousal (key in the evaluation and regulation of cognitive processes), they have a particular and a formal object (a cognitive process and its properties, respectively) and they motivate behaviour (both physical and mental). As way of example, let us take disorientation (i.e. a metacognitive feeling of low
— 95 —
confidence in the subject’s active navigational process) in its prototypical form (i.e. with negative valence and high arousal, as explored in chapter 2 and most of the cases in our corpus): The particular object of disorientation is the subject’s active navigational process, which is evaluated in a negative light, resulting in the formal object of low confidence, negative valence, high arousal and the appropriate behaviour to regulate the cognitive process in question, such as backtracking (physical action) or imagining a mental map of the area (mental action). Now that we have delineated the central features of metacognitive feelings, let us return to AIT in order to account for them within the PP framework. If we were to simply equate cognition with information-processing (as in Neisser 1967), then all feelings would be metacognitive, because (according to AIT) they evaluate information processing (i.e. prediction success; in the form of Expected Rate). Thankfully, the PP framework provides a more rigorous way to frame the idea of the cognitive domain traditionally conceived. The hypothesis-testing brain of PP operates in various predictive modalities, i.e. there are various dynamic areas in need of modelling and prediction, of which the world and the body are prominent ones. Each area will have their own deep hierarchical generative models tailored to the specifics and regularities of the domain. For instance, exteroception and interoception with their differentiated modalities grossly comprise the sensory domain that is concerned with modelling and predicting the external world and the body, respectively. The suggestion is that we can understand the cognitive domain in a similar way. But what is it that is modelled by classical cognition? In Metzinger's words: “[W]hat parts of the world can be accessed by neither exteroceptive nor interoceptive predictive processing? (…) The brain itself; the neural body” (Metzinger 2017, p.16). Following Metzinger’s line of thought, we should consider that the cognitive domain is best understood as the effort of the brain to predict its own activity. What belongs to the cognitive domain are those kinds of cognition that happen when the brain predicts itself. An important insight that connects AIT with empirical research on meta-cognition concerns the role and nature of processing fluency. The sub-personal property of processing fluency (or disfluency) is often considered to be the main proximal cause of positively (or negatively) valenced metacognitive feelings (e.g. Winkielman et al 2003). In PP terms, processing fluency can be reconsidered in terms of the Expected Rate associated with a cognitive process, because if (in line with fluency accounts of valence) the organism monitors the fluency of the information-processing of a given cognitive activity, changes in said fluency will result in changes in prediction success, so that if there is an increase in fluency the organism can expect a ceteris paribus proportional increase in Rate. According to AIT, a particular object corresponds to the inferred cause of expected changes in Rate and a formal object corresponds to a model of expected changes in Rate. In the case of metacognitive feelings, the cognitive process that the feeling is tracking will emerge as the particular object of that feeling, because its performance is expected to cause changes in Rate. How those changes in Rate will develop (and how the individual should respond) is what is modelled through the formal object of the metacognitive feeling. In the case of the FoK, expected success in retrieving a certain information (e.g. the answer to the question “what is the capital of Nepal?”) will lead to an increase in Expected Rate and to the experience of positive valence directed at the process of retrieving (i.e. the process required to answer the question) in the form of a formal object (i.e. a model that predicts imminent information retrieval; corresponding to the experienced sense of knowing the answer to the question). The subject would thus take her FoK not only to be about recalling the capital of Nepal (particular object) but also about her knowledge of it or her ability to recall it (formal object). Note that in the subject’s phenomenal experience, the particular object of a metacognitive feeling can be a process (answering the question) or the content at which that process is directed (the capital of Nepal, which appears as about to be revealed). More often than not it will be the content that the processes generate or carry that will appear to the subject as the particular object of the feeling. This is because the content is usually consciously accessible, and the processes
— 96 —
are not (Metzinger 2003). An interesting open question is whether under normal conditions a subject with sufficient conceptual sophistication might actually gain introspective access to her cognitive processes as cognitive processes (cf. Mandik 2006). Sometimes, the subject might not be (immediately) aware of the particular or formal object of her feeling because the processes and properties that give rise to it are not consciously available or the system fails in automatically identifying and binding the affective components to appropriate particular and formal objects (i.e. the sources of expected changes in Error dynamics). In some cases, this will result in reiterations of the identification stage, possibly on a high level. In other words: sometimes the subject will have to resort to conscious interpretation to specify her feelings. Concerning this issue, there is substantial empirical evidence that points to the importance of the perceived context in determining the nature of metacognitive feelings (Koriat et al. 2004). For example, in a learning experiment, framing questions in terms of forgetting (in contrast to remembering) reduced the subjects’ confidence (Finn, 2008). Interpreting these results from the AIT viewpoint, the way the subject frames the situation changes the resulting metacognitive feeling. If the subject directs her attention to forgetting, all of the elements that might support the hypothesis that the subject is forgetting something will have a higher weight when the system generates a model of Error dynamics, making the emergence of a FoF more likely. In contrast, if the subject directs her attention to remembering, the elements that might support the hypothesis that the subject will remember will have a higher weight when the system generates a model of Error dynamics, making the emergence of a FoK more likely. As with all affective experiences, valence and arousal will constitute the core affect of the feeling, and the particular and formal object will emerge (at a relatively slower pace) to frame (i.e. model) this core affect in a dynamic fashion. As we saw in section 6.3, AIT understands formal objects not only as descriptive models of Error dynamics, but also as action policies. The result is that when feelings emerge, they transform the subject’s affordance landscape and influence the subject’s navigation through said landscape. As for metacognitive feelings, we know that they influence not only physical but also mental action. Regarding mental action, Thomas Metzinger argues that possible mental actions (affordances to mentally act, i.e. the cognitive domain equivalent of proprioceptive predictions) are also processed in parallel and compete for further processing, creating in the process a cognitive affordance landscape (see Metzinger 2017 for a detailed PP account of how the autonomous activity of the mind wandering network creates and biases this landscape). The analogy between Cisek’s affordance competition hypothesis and Metzinger’s cognitive affordance landscape makes the understanding of how metacognitive feelings affect mental behaviour fairly straightforward. The formal object of a metacognitive feeling transforms this cognitive affordance landscape and biases the competition between different cognitive affordances. In other words, a metacognitive feeling makes certain mental actions (e.g. instances of remembering, calculating, reasoning or imagining) more probable than others. The idea that cognitive feelings express mental or cognitive affordances in a non-conceptual way has been defended before in the literature (Proust, 2009), and AIT gives us a framework to clarify how this works: cognitive feelings bias cognitive affordance competition, which in turn influences the probability of occurrence of different mental actions. To illustrate, picture a situation in which a subject is asked what the capital of Nepal is. The subject doesn't immediately come up with an answer but a FoK emerges. This FoK modifies the subject’s cognitive affordance landscape, making it more likely for the subject to undertake mental actions related to retrieving the name of the capital of Nepal (see also Michaelian 2012; Arango-Muñoz 2013). A lot of different actions would be positively biased in this way, such as imagining where Nepal is on the map or remembering the capitals of different countries in the region, or more simply, maintaining the cognitive process of information retrieval. When it comes to the system making predictions to mentally act, these possible actions will have become more likely to be selected and to be allocated the cognitive resources necessary to perform the mental actions. This
— 97 —
biasing of cognitive affordances happens at different levels and timescales. It is likely that not only immediate mental actions will become more probable because of cognitive feelings, but that distal predictions such as goals will become more probable as well (see Pezzulo and Cisek 2016). Back to our example, the FoK will make the hypothesis that the subject knows what the capital of Nepal is more likely, and the goal of retrieving the name of the capital of Nepal more likely to be pursued. It is important to emphasise that the dynamics involved here are non-linear. Because all of this happens in a feedback loop, metacognitive feelings will modify the likelihood of mental actions, but mental actions will also modify metacognitive feelings in turn, by modifying the expected Error dynamics and with them, the predictions (e.g. valence) determining the metacognitive feelings. In general, engaging in (mental) actions spans predictive trajectories in whose course feelings can arise in a regulative fashion (Proust 2013). What this means is that in performing the component steps that implement a mental action, Error dynamics unfold, leading to possible alterations in Expected Error, giving, in turn, rise to regulative (metacognitive) feelings in the form of predictions about Error dynamics. If, guided by the FoK, the subject spends a long time invested in mental activity trying to remember the capital of Nepal unsuccessfully, the hypothesis that the answer will be retrieved (and thus, the goal of retrieving the answer) becomes less likely. This makes the Expected Rate go down (because there is mounting evidence that the subject does not actually know and that information retrieval will be unsuccessful), making the positive valence diminish, and eventually making the FoK dissolve and possibly transform into a ToT experience with negative valence. Conversely, if the subject remembers that the capital of Nepal is Kathmandu, the FoK will also eventually dissolve, because the predicted increase in Rate (related to the process of memory retrieval) will have taken place. The expected epistemic gain that the FoK was singling will have happened. Error will have gotten reduced upon finding that the capital of Nepal is Kathmandu, thus fulfilling the goal and the corresponding prediction of the subject knowing the answer to the question. As no further Error reduction is expected related to memory-retrieval, the system will stop predicting related positive Rate and the FoK will dissipate.
VI – CONCLUSION The previous chapter started by outlining the characteristics of affective experiences which a PP theory should aim to account for. This chapter then showed how AIT provides a satisfactory account of the characteristic features of affective experiences: that they have phenomenal valence, that they have a certain degree of felt arousal, that they are intentional states with a particular and a formal object, and that they motivate behaviour. According to AIT, phenomenal valence corresponds to an agent’s prediction of her own success in modelling the world (Expected Rate). Felt arousal corresponds to an agent’s prediction of the rate of change of that success (Expected Change in Rate). Through the tracking and prediction of regularities, valence gets experientially bound to a particular object (the inferred cause of expected changes in Rate) and a formal object (a model of expected changes in Rate). Formal objects are better understood not as passive properties of the particular object, but as representations that are both descriptive and directive. Affective experiences favour certain behaviours, and the function of formal objects is to model the particular object and its link to Expected Rate so that action can emerge in a regulative fashion — enabling feelings to fulfil their role of guiding the organism to homeostasis and to optimal levels of uncertainty in order to minimise prediction error over time. In the final section, AIT provided an account for how metacognitive feelings emerge and guide behaviour, a necessary step towards this monograph’s overarching goal of providing a unified theory of disorientation. According to this account, the phenomenal valence and felt arousal of
— 98 —
metacognitive feelings correspond to the Expected Rate and the Expected Change in Rate associated with a cognitive process, respectively. The cognitive process in question corresponds to the particular object that is inferred to cause changes in Error dynamics. The performance of this cognitive process is modelled through a formal object, which is both a descriptive model and a directive action policy. Through this formal object, the metacognitive feeling guides action (both mental and physical) by transforming the subject’s (cognitive) affordance landscape and biasing the competition between potential actions.
— 99 —
7 . A F F E C T I V E S T A T E S A N D D I S T R I B U T E D
C O G N I T I O N
This chapter takes spatial disorientation as a case study to elucidate the role that affective states play in distributed cognitive processes. Section 7.2 reviews recent situated approaches to emotion. Section 7.3 introduces some cases of distributed cognition in navigation. Section 7.4 looks at the case of spatial disorientation to explore the role of affect in navigation within the broader context of situated cognition. In section 7.5, I argue that, in distributed cognitive processes, affective states serve as a form of evaluative regulation that contributes to the modulation of a distributed cognitive process and to the eventual switch between distributed cognitive processes.
I – DISTRIBUTED COGNITION Distributed cognition is a theoretical framework developed under the idea that all instances of cognition can be understood as emerging from distributed processes (Hutchins, 1995, 2001). Distributed cognitive processes may include elements at many spatial and temporal scales (Hutchins, 2014). In a liberal construal, both the brain in itself (e.g. the interaction of different brain areas) and the interactions of the brain and the body can be seen as distributed cognitive processes. On a larger scale, the interactions of humans with cognitive artefacts (e.g. in the use of a calculator) constitute an instance of distributed cognition, and, last but not least, a distributed cognitive process may also emerge from the interaction of several human agents. At an even larger scale, cultural practices within a given cognitive ecosystem (e.g. the emergence of language, as in Hutchins and Johnson 2009) can also be explored through the lens of distributed cognition. A paradigmatic example of a distributed cognitive process is the cultural practice of queuing, in which the emergent spatial arrangement of the agents’ bodies maps the order of arrival to the queue. This involves a cooperative social practice and often also a practice involving material objects, such as the presence of ropes indicating how the queue should form or a line on the floor indicating the appropriate distance between the first and the second person on the queue. What is particularly interesting in the case of queuing is the mental practice of seeing a physical structure (the linear arrangement of bodies) as a conceptual structure (the order of arrival), which affords certain cognitive inferences (e.g. order of arrival, estimated waiting time…). Seen in the light of distributed cognition, cultural practices such as queuing are forms of dimensionality reduction (i.e. a two-dimensional array of people on a surface is constrained into the approximation of a one-dimensional line) that increase the predictability of the situation (e.g. computing who will be next in accessing a service). Importantly, as Edwin Hutchins points out, this dimensionality reduction does not take place inside any one individual’s mind, but in the space shared by the participants of the practice (Hutchins 2005).
— 100 —
Queueing is an example of how humans turn material patterns into representations by enacting their meanings (Hutchins 2010). By enacting the cultural practice of queuing within the appropriate cultural context (i.e. a situation involving waiting within a society in which queuing is common), the people in line produce the phenomenal vehicle of interest and the queue’s physical array becomes an enacted representation of the order of arrival. Hutchins claims that enacted representations are dynamic (involving both memory and anticipation), multimodal (e.g. combining visual perception and bodily action), and “saturated with affect” (p. 434, Hutchins 2010). To see what Hutchins has in mind, it is useful to look at the example he provides from the use of manual plotting tools during ship navigation in a U.S. Navy ship in the early 1980s (Hutchins 1995). In normal conditions, fixing the ship’s position requires measuring the bearing from the ship to at least three landmarks. The navigators plot a line of position (LOP) with respect to each landmark using a special tool called “hoey”. The intersection of three LOP’s on a chart forms a triangle within which the ship is assumed to be located. A large triangle indicates that there are problems with the plotting process, and a small triangle indicates that the process is reliable. On one occasion, the main gyrocompass of the ship failed, and the crew became unable to follow the practice described above. They were required instead to calculate the true bearing of a landmark, which equals compass heading plus magnetic deviation (the errors of the compass due to the local magnetic environment) plus magnetic variation (the difference between the direction of the earth’s magnetic field and true north from a given point) plus the relative bearing of the landmark with respect to the ship’s heading. The crew had to try to figure out how to calculate true bearing while plotting thirty-eight lines of position. They were missing the deviation term from their calculations and the three lines of position resulted in unusually large triangles. Eventually, the plotter realised what was missing, and once he included the deviation term, the crew gradually became able to fix the position of the ship again. What is interesting about the failing gyrocompass incident is the process through which the plotter realised that the deviation term was missing. The discovery came through an “Aha!” insight obtained from the plotter’s bodily engagement with the plotting tools through enacted representations. As mentioned earlier, the resulting triangles were at first unacceptably large. The plotters exclamation “I keep getting these monstrous frigging god-damned triangles!” gives us an impression of how strongly emotional the process was. During the whole process, the plotter shifted his own body around (e.g. placing his index finger on the location of the landmark) and moved the chart and the hoey to imagine LOPs that would make the fix triangle smaller. This continued until the elements of the involved enacted representations combined so that the plotter was able to imagine a small clockwise rotation superimposed on his visual experience of the protractor scale, leading to the realisation that adding a small number to the bearing for LOP3 would reduce the fix triangle. After this moment of “Aha!” insight, the plotter understood that what had been missing from the calculations was deviation (3º), which would improve not only LOP3, but all the LOPs, resulting in the desired small triangle (Hutchins 2010). The above is an excellent example of distributed cognition because the necessary calculations to fix the ship’s position did not take place solely inside the plotter’s head through internal mental representations, but instead took place through enacted representations within a cultural ecosystem. Such enacted representations are embedded in durable material media —what Hutchins refers to as “material anchors for conceptual blends” (Hutchins 2005)— and in bodily processes (somatic anchors), producing a multimodal integration that, if congruent, leads stability to the enacted representations employed in a distributed cognitive process. It is important to note that for a distributed cognitive process to emerge, the necessary multimodal integration of different elements often involves more than one agent. And as Hutchins points out, affect thereby plays an important role in distributed cognitive processes, but what this role is precisely is yet to be clarified in the literature.
— 101 —
II – SITUATED APPROACHES TO EMOTION Situated approaches to emotion offer an alternative to a long tradition of considering emotions as purely internal states or processes. In contrast to such a tradition, situated approaches consider emotions as forms of skilful engagement with the world that are both scaffolded by and dynamically coupled to the environment (Griffiths and Scarantino 2005). A tenet coming from this perspective is that emotion is designed to function in a social context and “is often an act of relationship reconfiguration brought about by delivering a social signal” (p.2, Griffiths and Scarantino 2005). To illustrate the degree to which emotion is designed to function in a social context, Griffiths and Scarantino draw on paradigmatic cases of situated emotion such as social appraisal (Sorce, Emde, Campos, & Klinnert, 1985) or audience effects (Fernández-Dols & Ruiz-Belda, 1997). Social appraisal cases are those in which an individual’s appraisal of a situation relies on that of others. Social appraisal is especially interesting for our study of distributed cognition because it shows how the appraisal of a situation is often distributed beyond the individual. Emotions play an important role in these distributed cognitive processes. A well-known example is an experiment showing that the willingness of infants to crawl over a visual cliff was influenced by the negative or positive facial expressions of their mothers (Sorce, Emde, Campos, & Klinnert, 1985). Instances of social appraisal are very common, ranging from judging if a situation is dangerous to judging if a joke is funny or offensive. Duchene smile (involving both the movement of the mouth and of the eyes — generally interpreted as a display of happiness) is a good example of audience effects. When bowling, whether bowlers smile or not after knocking down some pins depends much more on whether or not they are facing their companions than on the actual number of pins they knock down, so that they will often smile a Duchene smile after knocking a few pins while facing their companions, but not smile after a full strike while facing away (Kraut & Johnston, 1979). Similar effects are found in football fans and even in Olympic medallists (Fernández-Dols & Ruiz-Belda, 1997). Audience effects point to emotions being used in a transactional way, rather than being primarily internal states. The understanding of embarrassment as a sign of acceptance of social norms one just infringed upon (Leary, Landel and Patton, 1996) or the interpretation of sulking as a strategy of relationship reconfiguration (Parkinson 1995) also point to the transactional nature of emotions. In turn, emotional management (which, as with social appraisal, is often distributed beyond the individual) influences the unfolding of the emotion itself, which sustains the situationist idea that emotions are dynamically coupled to the environment, insofar as they work in “a feedback mechanism which involves the reciprocal exchange of signals delivered by expressions and other behaviour in the course of time” (Griffiths and Scarantino 2005, p. 19). An important distinction to be made when considering the environmental scaffolding of emotions is that this scaffolding happens both synchronically (in the unfolding of the emotion itself) and diachronically (in the process of acquiring an emotional repertoire). In any given society, the environment scaffolds the development of the appropriate emotional repertoire diachronically through ideational factors such as normative standards, shared expectations (about the unfolding of emotions) or belief systems (about the nature of emotions) and material factors such as venues with specific emotional affordances (e.g. a theatre favours displays of elation and a bar favours displays of joy) or emotional technologies ranging from depression medication to stress balls (Parkinson, Fischer and Manstead 2005). Synchronically, the environment scaffolds the unfolding of different
— 102 —
emotions through material elements (e.g. a confessionary box), rituals (e.g. the different stages of a wedding and the corresponding expected emotions) and social processes such as the afore-mentioned social appraisal. Through the lens of distributed cognition, emotions can be understood as cultural practices unfolding (i.e. synchronically scaffolded) within a particular cognitive ecosystem (i.e. diachronically scaffolded). Let us call this subset of cultural practices (i.e. practices in which emotions play a paradigmatic central role) emotion practices. Before going any further, some clarification is needed. The situated approach to emotions stays neutral about the stronger ontological claim that emotions extend onto the environment. The approach does not necessarily imply that the environment is a constituent part of an emotion; the claim is simply that the environment has a casual contribution to emotion. What the approach offers is a methodological alternative to traditional emotion accounts, just like distributed cognition offers an alternative to traditional accounts of cognition without making ontological claims about the extension of the mind. The synergy of the two theoretical approaches has a great methodological potential and it can help us address the central research question of this chapter, namely, the role of affect in distributed cognitive processes. We have already seen two ways in which distributed cognition can help the situated approach to emotions. One way is by providing a good framework for analysing cases such as social appraisal. The other is by driving the distinction between synchronic and diachronic environmental scaffolding of emotions into a reconceptualization of emotions as cultural practices within larger cognitive ecosystems. This insight can in turn help us rephrase our central research question: what is the contribution of emotion practices to the broader cultural practices of a given cognitive ecosystem? Answering this question requires us, on the one hand, to clarify the diachronic scaffolding of said cognitive ecosystem and, on the other, to analyse the synchronic scaffolding of emotion practices and its impact on distributed cognitive processes. To do the this, we further need to clarify the nature of the dynamic coupling of emotions with their environment. The control-process theory of affect offers a promising way of doing so. Carver and Scheier use a feedback-based viewpoint on self-regulation and the analogy of action control to explain the emergence and unfolding of affect. They argue that affect works as a feedback system to regulate the unfolding of action in an attempt to diminish distance to a goal (Carver & Scheier 1990, 2001). Proust takes the control-process view as a point of departure to argue that feelings, in their regulatory and evaluative role, “constitute the sensitive part of predictive and retrospective processes of nonconceptual evaluation of one’s own and others’ well-being and actions” (Proust 2015, p.4)25. We can then apply this broader control-process conception to the situationist insight that emotions often function as forms of relationship reconfiguration. In a social context, emotional behaviour lies on a spectre between expression (e.g. solitary laughter) and negotiation (e.g. ingratiating smile), so that certain emotional behaviours should be understood as a form of negotiation rather than as the expression of an existing inner emotional state (Hinde, 1985, p. 989). In a social context, the evaluative aspect of emotions is made public through expression, and the regulative aspect of emotions underlies the negotiation side of emotional behaviours. Together, these two aspects make it possible for emotional practices to work as feedback mechanisms for the dynamical and distributed reconfiguration of relationships.
25 Chapter 6 provided an account of the evaluative-regulative role of affective states within the Predictive Processing
framework. Although the two frameworks are compatible, it is only in the final chapter that I combine the two accounts in order to build a theory of spatial disorientation.
— 103 —
III – NAVIGATION From fixing the position of a Navy ship to following a well-marked trail, navigation offers some of the best examples of distributed cognitive processes. It often involves cultural practices such as collective decision-making (e.g. chart plotting), social appraisal (e.g. relying on a guide or a group), environmental scaffolding (e.g. trail signage, cairns…) or the use of cognitive artefacts (e.g. compass, GPS devices), all of which need to take place within the appropriate cognitive ecosystem. So far, we have seen how a given cognitive ecosystem provides the scaffolding for emotion practices. Now we turn our focus to the other side of the coin. The question at hand is how emotions provide the scaffolding for broader cultural practices, such as navigation. Here again, the scaffolding can happen both diachronically and synchronically. Diachronically, emotion practices work as ideational factors that promote and facilitate the appropriate cultural practices within a cognitive ecosystem. In his longitudinal study of the social dimensions of geographic disorientation in Arctic Alaska, Sonnenfeld highlights the disappearance of the practice of shaming as one of the causes underlying the increase in disorientation episodes among the Inupiat residing in Wainwright (Sonnenfeld, 2002). It turns out that in the 1960s there was a widespread practice of shaming the individuals who happened to get lost in the wild. The emotion practice of shaming involves the shaming of the individual on the part of the community, the embarrassment of the individuals who get lost and, most importantly, the fear of such potential embarrassment, which incentivises the learning and application of the appropriate navigational practices: "Fear of such ridicule forces the Eskimo to learn his navigation skills well and to exercise caution whenever he travels" (Nelson 1969, p. 386). By setting the normative standards, the emotion practice of shaming contributed to the diachronic scaffolding necessary for the development of the appropriate cultural practices (e.g. navigational skills) of the Inupiat cognitive ecosystem in Arctic Alaska. Consequently, the disappearance of this emotion practice after the 1960s resulted in a degradation of the broader cultural practices of navigation and in the subsequent increase of disorientation episodes that Sonnenfeld witnessed in the early 2000s. Sonnenfeld’s study can also offer us an insight into how emotion practices can provide scaffolding for navigational practices synchronically. Travel among the Inupiat is often group travel, which involves a great deal of relationship reconfiguration. An advantage of traveling as a group is that it allows for the distribution of cognitive tasks. When hunting, one of the members of the group can follow an animal trail while another member focuses on staying oriented, so that the group as a whole is not lost after the hunt. Even without a clear division of tasks, traveling in a group can facilitate orientation because if one of the members fails to orient correctly, there is a chance that another member of the group has maintained the correct orientation. Nevertheless, this can only work if the group can determine whose orientation is correct when there is a conflict. When two members of a group disagree on the direction that the group should follow, relationship reconfiguration is needed, and this is a highly affective process involving confidence, doubt, trust, mistrust, anger, pride and even fear. During navigational group conflict, emotion practices provide the synchronic scaffolding to evaluate and regulate distributed cognitive processes. Confidence and doubt arise to evaluate the ongoing distributed cognitive process of orientation (both one’s own and that of one’s companions) and to regulate it as well (by continuing the process, adjusting or terminating it). Feelings such as anger can be used to reconfigure the existing relationship by trying to force the group to follow one’s sense of orientation rather than that of another member. And mistrust can even force the dissolution of the group, so that each member can follow the way they see fit. The example of Inuit group travel brings to the fore a picture in which the synchronic scaffolding of emotion practices works as a way of evaluating and regulating a distributed cognitive system. In the following sections, a phenomenological analysis of disorientation will clarify this picture.
— 104 —
IV – DISORIENTATION AS AN AFFECTIVE STATE In chapter 4, I argued that disorientation should be characterised as a metacognitive feeling that serves to evaluate and regulate active navigational processes. This characterisation helps us make sense of the phenomenology of disorientation presented in chapter 2: via affordances, we can understand the link between this regulative-evaluative process and the degradation of possibility that subjects undergo during spatial disorientation, because feelings (by regulating and evaluating changes in the relation between an organism and its environment) serve to monitor affordances, making them more or less salient (Proust 2015). Disorientation, by evaluating orientation, transforms the saliency of the surrounding affordances, which is subjectively experienced as a diminished sense of possibility, inducing the corresponding alienation from the immediate environment. While disorientation dims down certain affordances, it makes other affordances more salient. Most often, disorientation pushes one to reconsider one’s orientation, prompting cognitively demanding actions such as revisiting one’s mental image of the spatial layout. More technically, disorientation is known to elicit a switch from an online system of spatial representation to an offline system of spatial representation (Waller and Hodgson 2006). The offline system of spatial representation produces enduring and stable spatial representations of the environment. In contrast, the online system produces transient representations that are dynamically updated through spatial perception, i.e. it integrates egocentric and allocentric frames of reference. The experimental finding that subjects switch from online to offline systems of spatial disorientation when disoriented is very much in line with its characterisation as a metacognitive feeling. The feeling of disorientation assesses the online system to be unreliable (this is its evaluative function) and elicits the switch to the offline system (this is its regulative function). Using the offline system, the subject tries to consciously re-evaluate the alignment of the allocentric and egocentric frames of reference of the surrounding environment. In navigation contexts, this process is often distributed beyond the individual and involves the use of maps, compass, navigation instruments, GPS technology or environmental cues. If the source of the error is resolved (e.g. Binet’s subject realising that he had turned on the wrong direction, see chapter 2), the effect is often an “aha moment” (as in Hutchins’s case of the plotter finding the correct deviation) and a sudden phenomenal realignment follows (e.g. the allocentric frame of reference of a turned-around subject switching 180º to the correct orientation). Reconsidering one’s orientation is not the only action that is prompted by disorientation. A common reaction of people who get lost in the wilderness is exploration. Kenneth Hill conducted over a hundred interviews with subjects who had become lost in the wilderness and identified a series of common strategies in lost person behaviour (see chapter 1). Some of these strategies were driven by a desire to reach civilisation as quickly as possible (e.g. following a single route for as long as it takes), but many of them were based on exploring the surroundings to facilitate reorientation. Three clear examples of disorientation-induced exploration are route sampling, direction sampling and route enhancing. Route sampling consists in using an intersection as a base for exploring different routes. Direction sampling consists in using a visible landmark as a base for exploring different directions. View enhancing consists in aiming for a high position in order to gain visibility. Additionally, disorientation can be highly arousing, and it is known to push subjects into panic, which results in suboptimal behaviours such as random walking (Hill 1998).
— 105 —
Before closing this section, it is important to note two things. The first is that disorientation is not a discrete, on-or-off affective state. It comes in degrees, and it sometimes nudges navigational behaviour rather than prompting a complete reassessment of the subject’s orientation. When one gets turned around, disorientation is a very sudden, arousing and salient feeling. However, when one is following a map through a city, a slight feeling of disorientation might simply prompt one to double-check that we are still on the street we are supposed to be following. The second thing to note is that orientation itself is also an affective state. This becomes clear when we look at the being-turned-around cases: once one understands what’s going on and the spatial representation switches 180º to its correct position, one feels oriented. Of course, orientation does not often come to the fore because most of the time it is a background feeling. Still, we should consider the affective dimension of navigation as a continuum between orientation and disorientation, helping us guide our navigational activities in an evaluative-regulative fashion.
V – NAVIGATION, AFFECT AND DISTRIBUTED COGNITION All instances of cognition can be understood as emerging from processes distributed at many spatial and temporal scales, and navigation is no exception. Navigation is performed in a particular cognitive ecosystem (e.g. involving signage codes, trail conventions, adapted navigational skills…) and is distributed within individuals (involving processes of attention, memory and decision-making) and beyond individuals (involving several group members and cognitive artefacts such as compasses and maps). Consequently, when disorientation evaluates and regulates navigation, it is evaluating and regulating a distributed cognitive process. If a map is indicating a route that no longer exists, if a malfunctioning GPS is misrepresenting one’s heading or if a demagnetised compass is failing to point north, and one becomes disoriented as a result, the affective state of disorientation is not simply evaluating an internal process, it is evaluating a distributed cognitive process that extends beyond the individual. Inseparable from this evaluative aspect is the regulative aspect of disorientation, which can range from modulation (e.g. turning the map around in an effort to elicit alternative interpretations; or restarting the GPS) to disengagement (e.g. pocketing the compass and trying to infer north from the position of the sun)26. The notion that disorientation often leads to disengagement resonates with dual process accounts of spatial cognition (see Allen and Haun 2004 for a review), and in particular with the idea that disorientation can cause a switch from an online to an offline system of spatial representation (Waller and Hodgson 2006). This online-to-offline switch is a switch from a dynamic, fast operating system to a slow system modulated by will and attention, which is in line with the distinction between two modes of operation of the brain in certain tasks; M1 being automatic, dynamic, fast autonomous and stimulus-driven (exploiting heuristics and thus prey to biases; see Tversky and Kahneman 1974), and M2 being slow, heavily-reliant on working memory and modulated by will and attention (Evans 2003 and 2012 for reviews; Evans and Frankish 2009; Kahneman 2011). Consequently, disorientation can lead to a switch from an M1 (online) to an M2 (offline) system of spatial representation.27 In a recent account, Roberto Casati expands the M1-M2 distinction to account for the use of cognitive artefacts (Casati 2017). Although much of the literature has focused on cognitive
26 Another variation of disengagement is the initiation of contingency measures, such as the co-pilot starting to double-
check the route on their own phone once they feel that the driver has lost the way. 27 As mentioned in chapter 4, a notable exception is that highly stressful disorientation episodes might push the subject
towards M1, as stress is “a state in which an individual is unable to instigate a clear pattern of behavior to remove or alter the event (…) that is threatening an existing goal” (pp. 177, Power & Dalgleish, 2007).
— 106 —
offloading —the full delegation of the mental activity onto a cognitive artefact (see Risko and Gilbert 2016 for a review),— which Casati dubs the M4 operating mode, there is also an intermediate level of delegation between M2 and M4, which Casati terms M3. In M3, the subject’s interaction with a cognitive artefact is essential to the cognitive task, but it does not involve a wholesale offloading of the core computations, as is the case with M4. In navigation, an example of M4 is the use of a GPS that gives us directions to a goal, and an example of M3 is the use of a compass and a map to find our way. M4, like M1, is fast and stimulus driven. M3, like M2, is slower and requires will and attention. A recent experiment shows that spontaneous metacognitive assessments of confidence trigger cognitive offloading (Boldt and Gilbert 2019). In the terms of our present discussion, this can be understood as a feeling of confidence triggering a switch from M2 to M4 (see Fleming and Dolan 2012 and Yeung and Summerfield, 2012 for reviews of the feeling of confidence). In Waller and Hodgson’s experiment, disorientation is evaluating an internal system of spatial representation to regulate spatial cognition (i.e. the switch from M1 to M2). In Boldt and Gilbert’s experiment, a feeling of confidence is evaluating internal memory processes to regulate the subject’s level of engagement with a cognitive artefact within a distributed cognitive process (i.e. the switch from M2 to M4). Although Boldt and Gilbert did not test for a feeling of confidence directed at the cognitive artefact, it is logical to assume that if the involved cognitive artefact were evaluated as unreliable, the inverse process would be triggered, and the subject would disengage from that distributed cognitive system (i.e. a switch from M4 to M2). We can expect that something similar is going on regarding M3. If an agent feels disoriented, she might pull out her map to aid navigation (i.e. a switch from M1 to M3). If later on, the map does not seem to match the street network, this might provoke a switch from M3 back to M2 or M1 (e.g. trying to picture a mental map of the city or trying to backtrack until she reaches a familiar landmark). Here, there is no reason for the reader to buy wholesale a strong, extended version of the dual processing account. What is key is that affective states can elicit a switch between two cognitive processes, and that these cognitive processes are often distributed at different spatiotemporal scales and often have substantial different characteristics in terms of automaticity and speed, of working memory use and of responsiveness to endogenous attention. At this point, we already have all of the elements of the present account. According to the account, affective states evaluate and regulate distributed cognitive processes both diachronically and synchronically. Diachronically, this can lead to the promotion of certain cognitive processes (e.g. shame promoting the learning of navigational skills). Synchronically, the evaluation of different modes of cognition distributed at different scales (M1-M4) can lead to modulation (of a single cognitive process, and therefore within one mode) or disengagement (i.e. switching between cognitive processes, and most often between M1, M2, M3 and M4 modes of processing). As I mentioned in the closing of the previous section, affective states are not discrete, so it might be that the finding of switches rather than transitions between modes is partly a result of experimental settings. I expect that in certain situations, mild affective states (e.g. of disorientation or orientation) lead to different levels of engagement rather than to on-off disengagement (e.g. double-checking street names if we are not that confident of our GPS system, rather than turning off our GPS), and that in most cases mild affective states lead to the modulation of distributed cognitive processes in ways that might not be very apparent. Most of the time, when navigating in a group, if a member of the group feels disoriented, this does not lead to a burst of anger that breaks the group apart, but to shows of doubt (e.g. forms of emotional negotiation such as unsure smiles or raised eyebrows) that make the expedition leader double-check that the expedition is well-oriented. Before concluding, I would like to link the present account of affect and distributed cognition back to the phenomenology of disorientation as was presented in chapter 2, namely to the sinking away
— 107 —
of affordances and the ensuing alienation. I have already mentioned that affective states serve to monitor affordances and that they become more or less salient as a result. It is easy to see how disorientation, qua affective state, results in the sinking away of affordances (in particular, navigation-related affordances). What is interesting here about the novel account introduced in this chapter is that we can extend this understanding of disorientation to distributed cognitive processes. When a subject is engaged in a distributed cognitive process involving either other subjects or cognitive artefacts, the disengagement following from disorientation affects the affordances that were facilitated by the particular distributed cognitive process. If the different elements of the cognitive process (e.g. material media and somatic anchors) required to produce enacted representations fail in their multimodal integration, the enacted representations disintegrate as a consequence, and with them the corresponding affordances. If we realise that the map that we have been following has led us to the wrong trail, in our disorientation we become alienated from both the map, which stops representing our surrounding space, and the trail, which is no longer a meaningful trail to our destination. If we become disoriented after following our GPS, the GPS becomes conspicuous, as the navigational potential it afforded sinks away. Equally, we become alienated from the other members of a group if we stop feeling like a member of that group. Multimodal integration within a distributed cognitive process leads to the emergence of (often interpersonal) phenomenal objects. Consequently, when a negative affective state triggers the disengagement from that distributed cognitive process, the affordances sustained by it sink away and the subject becomes alienated.
— 108 —
8 . A U N I F I E D T H E O R Y O F D I S O R I E N T A T I O N
The central claim in this monograph is that disorientation is a metacognitive feeling. The aim of this chapter is to build on this claim to develop an account of how disorientation emerges and how it guides behaviour. In order to do so, I will provide the necessary connections between the other chapters in the monograph. In section 8.1, I will provide a synthesis of PP and Distributed Cognition. In section 8.2, this synthesis will be used to expand the Affective Inference Theory (AIT) introduced in chapter 6. Section 8.3 will apply AIT to the metacognitive feeling of disorientation. Section 8.4 will explore how the emergent unified theory of disorientation appertains to the phenomenology of disorientation introduced in chapter 2. Section 8.5 will revisit the cases of the positive aspects of disorientation and of Evenki navigation (introduced in chapter 3) in light of the new theory of disorientation.
I – PREDICTIVE PROCESSING AND DISTRIBUTED COGNITION As we saw in chapter 5, according to PP, the organism will try to reduce Error over time by updating predictions and by changing patterns of precision gain, which involves heuristics (e.g. Optical Acceleration Cancelation), passive bodily dynamics (e.g. bipedal motion) and cultural practices — thus distributing work between brain, body and world. This approach shows how Predictive Processing and Distributed Cognition can work as separate but strongly compatible theoretical frameworks, something explicitly acknowledged by major proponents of both frameworks (see Hutchins 2013 and Clark 2015 for prime examples). In this section, I will outline a substantial challenge to the bridging of PP and Distributed Cognition —the problem of entropic cultural practices— and provide a solution to said challenge. The key here is the insight that cultural practices tend to decrease entropy (and thus, increase predictability, because entropy is a measure of unpredictability) across all scales in a cognitive ecosystem (Hutchins 2013). The work of Edwin Hutchins provides a comprehensive (but probably not exhaustive) inventory of the ways in which cultural practices achieve this (Hutchins 2012), which include:
• dimensionality reduction — the production of a conceptual structure out of complex assemblages of possibly preconceptual material via the conjunction of features (e.g. the production of a queue out of a group of humans via the conjunction of position in the line with order of arrival),
• filtering — preserving some features or elements while ignoring others (e.g. directing our attention to the white lines painted at the edge of a mountain road),
• constraint satisfaction — the simultaneous fulfilment of multiple restrictions that change the probability of different configurations of cultural practices emerging (e.g. when cycling
— 109 —
there is a simultaneous fulfilment of the constraints of the human body, the mechanics of the bike and a rich legal and cultural code, which makes certain ways of cycling more likely to emerge),
• modulated positive feedback — recycling a (filtered) subset of the output as an input (e.g. decreasing speed when our car gets dangerously close to the white line at the edge of the road),
• superposition of structure — the projection of imagined structure onto elements of a perceived or imagined world (e.g. seeing a constellation when looking at the stars).
• mapping across conceptual spaces — combining filtering with constraint satisfaction and superposition in order to map patterns from one conceptual space to another (e.g. translations, comparisons, analogies, metaphors…),
• and design — activities outside the normal workflow that attempt to create explicit representations of work practices (e.g. agreeing on a chain of command within an organisation).
• Hutchins discusses how these methods of entropy reduction are present not only in relatively straightforward cultural practices such as queuing, but also in highly intellectual and complex practices such as quantum physics research. To connect the above with PP, think again about the case of the baseball player (introduced in chapter 5) who uses Optical Acceleration Cancelation (OAC) to catch the flying ball. The player’s body needs to be dynamically positioned in such a way that the ball seems to move at a constant speed through the visual field. The way this happens is by ascribing high gain to the Error associated with cancelling the vertical acceleration of the ball’s optical projection and quashing other errors (Clark 2015). Following Hutchins’s inventory, OAC is an example of modulated positive feedback, a composite cultural practice that involves filtering in a dynamic loop. The baseball player filters the Error unrelated to OAC and dynamically updates position in order to minimise the changes in perceived acceleration. Thinking about cultural practices within the framework of PP, Kirchoff and Kiverstein argue that “the constraints that come from cultural practices influence how precision is weighted in a given context and thus how uncertainty is kept to a minimum” (p. 97, Kirchoff and Kiverstein, 2018). This is exactly what happens in OAC. The constraints of the cultural practice influence how precision is weighted (estimating high gain for OAC-related Error) in order to reduce uncertainty (about catching the ball). If we follow the principle of active inference, when we say that cultural practices tend to increase predictability, we are not talking only about the descriptive aspect of predictions, but also about their directive aspect. By engaging in OAC, the baseball player is not only making the prediction of where the ball is going to fall more reliable, but also making the prediction that said player will catch the ball more reliable. Cultural practices tend to increase predictability (or analogously, decrease uncertainty or entropy), but this is not the full story, because some cultural practices tend to decrease predictability instead (or analogously, increase entropy). Let us call this latter type ‘entropic cultural practices’, as opposed to the standard ‘negentropic’ cultural practices. This is a point that Hutchins himself acknowledges, although not one into which he delves in depth. Hutchins provides no inventory of entropic cultural practices, he simply notes that these exist, and that human cognition moves through cycles of disorder and reorder on all time scales. The only concrete example that he points to is that “in the conduct of a scientific investigation, accumulating disorder may lead to a productive conceptual reordering”, and he notes that “a jolt of unpredictability is sometimes needed to overcome stable but inadequate conceptual structures” (p. 321, Hutchins, 2012). It is important to take the time now to discuss negentropic cultural practices for two reasons. The first one is that they will be paramount to explain the positive aspects of disorientation in section 8.5. The second reason is that the existence of cultural practices that actually decrease predictability poses a problem both for Distributed Cognition and for PP: for Distributed Cognition, it poses a
— 110 —
problem because everything either increases or decreases entropy, so the connection between entropy and cultural practices becomes much less meaningful once we simply say that some cultural practices decrease entropy and others don’t. For PP, it poses a problem because we lose the elegant connection between increasing predictability and the PEM principle. To fully connect Distributed Cognition and PP, we need to address the problem of entropic cultural practices. Before we go any further, it is important to clarify that the problem of entropic cultural practices is not that a given practice does not increase predictability in some instances. Cognitive ecosystems merely increase the probability of certain patterns emerging, i.e. the interaction between the cognitive ecosystem and the resulting pattern of activity is non-linear. In these dynamic systems, it makes no sense to fixate on regular causation (i.e. causation in the sense in which if A causes B, then A must always be followed by B). A regularity notion of causation cannot be meaningfully defined for systems without linear interactions among their variables (for a mathematical discussion see Wagner 1999). When we are discussing cultural practices, we should follow a probabilistic notion of causality. Using this notion, to say that negentropic cultural practices increase predictability is to say that negentropic cultural practices make an increase in predictability more likely. The problem of entropic practices comes only when we find some cultural practices that tend to decrease predictability instead of increasing it. To solve the problem of entropic cultural practices, we need to bring our attention to how cognition is widely distributed both spatially and temporally. A first step in this direction is recognising how cultural practices enable learning. Games are a prime example: they provide us with an initial high entropy (e.g. we fail to answer the questions correctly in a trivia game), we slowly progress at reducing that entropy (e.g. we answer the questions with more and more accuracy) and we learn in the process (e.g. we improve our general knowledge). Learning is a point that Clark emphasises in his effort to bridge cultural practices and PP: “by designing and repeatedly redesigning our own environments, populating them with new books, paintings, theories, games, and practices, we humans continually move the goalposts for our own prediction-based learning” (p. 531, Clark 2018). The idea is that by being embedded in a culture, we learn to value (i.e. assign high precision to) certain cultural practices that help us (by enabling learning) in our goal of reducing Error over time. Nevertheless, practices that enable learning do not seem to be quite what Hutchins is referring to by “a jolt of unpredictability” and “accumulating disorder”. Hutchins was thinking in particular of scientific practices aimed at undermining existing theories. Another good example of such disruptive entropic cultural practices outside of the scientific context is the detournement technique of the Letterist International, which consists in hijacking the aesthetic elements of the dominant culture and transforming them in a subversive way (e.g. rearranging the letters in a slogan or a brand name) (Debord and Wolman 1956). Most protest practices also illustrate disruptive entropic cultural practices. The issue with these disruptive cultural practices is that they cannot be easily explained away by appealing to learning. There is no clear way in which they enable the gradual improvement of prediction success. To cover both the learning and the disruptive types of entropic practices, I will advance and defend the following conjecture: entropic cultural practices might decrease predictability in the short term and increase it in the long term. This is certainly what happens in the case of learning, and, I will argue, it is also what happens with disruptive cultural practices, such as the case of scientific investigation mentioned by Hutchins. A range of entropic cultural practices destabilise a theoretical structure (e.g. by coming up with problems that the structure cannot assimilate), which prompts a conceptual reordering. Disruptive practices are useful when cognitive ecosystems revolve around negentropic cultural practices that increase predictability in globally suboptimal ways. In order words, a given negentropic cultural practice survives because it manages to increase predictability,
— 111 —
but there is a latent possibility of new negentropic cultural practices that will do a better job at increasing predictability. Then, a likely role for disruptive practices would be to destabilise the cognitive ecosystem until a new negentropic cultural practice emerges that does a better job at decreasing predictability. The role of disruptive cultural practices is precisely to induce these periods of fruitful disorder28. Paying proper attention to the dynamics of cognitive ecosystems and to how cultural practices operate at different scales both spatially and temporally allows us to tackle the problem of entropic practices. All cultural practices tend to reduce entropy. The caveat is that different practices reduce entropy at different time scales. Negentropic cultural practices (like the ones in Hutchins’s detailed inventory) do so in relatively short temporal scales. Entropic cultural practices (disruptive practices and learning practices) do so at longer time scales. Together, they contribute to an increase of predictability over time in a given cognitive ecosystem. This increase in predictability over time contributes to the Error minimisation activities of the humans operating within said cognitive ecosystem. This characterisation of cultural practices provides the sought for connection between PP and Distributed Cognition.
II – AFFECTIVE INFERENCE THEORY ‘IN THE WILD’ In chapter 6, AIT provided an account of how metacognitive feelings emerge from the evaluation and regulation of a cognitive process and how they guide behaviour:
• Phenomenal valence corresponds to the Expected Rate associated with the cognitive process.
• Felt arousal corresponds to the Expected Change in Rate associated with the cognitive process.
• The particular object corresponds to the cognitive process that is inferred to cause changes in Error dynamics.
• The performance of the cognitive process is modelled through a formal object, which is both a descriptive model and a directive action policy.
• Through its formal object, the metacognitive feeling guides action (both mental and physical) by transforming the subject’s (cognitive) affordance landscape and biasing the competition between potential actions.
In chapter 7, I provided a parallel account of how affective states evaluate and regulate distributed cognitive processes both diachronically and synchronically. Diachronically, affective states can lead 28 Here, the question arises of why there would be suboptimal negentropic practices in the first place. There are two
reasons why suboptimal negentropic practices might emerge, and both are related to the context in which they develop. The first reason is path-dependency, a popular notion in historical sociology (Mahoney 2000): cultural practices develop in contingent ways, out of the vicissitudes of their epoch and environment. This might lead to practices that are just ‘good enough’ at a local level (for a defence of such a “good enough principle” in linguistic practices see Ferreira and Patson 2007). The second related reason is that negentropic practices might emerge in a particular situation in which they reduce entropy efficiently and then persist even when their efficiency has substantially decreased over time (e.g. think about how users might stick to computer programmes they know well even when better programmes have already been developed). These suboptimal practices persist because cultural practices are both self-reinforcing (e.g. the more people respect queues, the more likely people are to form queues) and mutually reinforcing (e.g. “the ways of speaking about ‘first in line,’ ‘next,’ ‘back of the line,’ and so on are discursive practices that enter into relations of mutual reinforcement with the conception of the linear spatial array as a queue” p.8, Hutchins, 2013), and, most importantly for our present discussion, because there might be no way of moving to a more efficient negentropic practice without first undergoing a period of decreased efficacy in terms of predictability.
— 112 —
to the promotion of certain cognitive processes (e.g. shame promoting the learning of navigational skills among Inupiat natives). Synchronically, the evaluation of different modes of cognition distributed at different scales (M1-M4) can lead to
• modulation (of a single cognitive process, and therefore within one mode), • or to disengagement (i.e. switching between cognitive processes, and most often between
M1, M2, M3 and M4 modes of processing).
The discussion in the previous section will clear the way to see how the approaches in chapters 6 and 7 come together. Let us look first at the diachronic role of affective states in the cognitive ecosystem. We saw that cognitive ecosystems provide constraints that make certain cognitive assemblies more likely to emerge. Here, background emotion practices (e.g. shaming) work as constraints that make certain cultural practices more likely to emerge (e.g. learning navigational skills). In PP terms, they promote certain patterns of precision weighting. Taking the case of the Inupiat, the expectation of potentially being shamed ensures that, ceteris paribus, the engagement with (entropic) learning practices and with (negentropic) navigational practices becomes the most effective way to reduce Error over time. Inversely, learning cultural practices will enable agents to finetune their affective machinery by learning what cues best predict Error dynamics and how to model those dynamics in ways that are adjusted to the particular cognitive ecosystem (e.g. what one should be afraid of, how one should express sadness, what to do when we mistrust someone…). We do not need AIT to understand how affective states constraint cognitive processes diachronically, because AIT is aimed primarily at the synchronic links between affective states and cognitive processes. AIT provides a way to understand how affective states evaluate cognitive processes distributed at different scales synchronically. When a cognitive process is ongoing at whichever scale of distribution, it is associated with certain expected Error dynamics. As a result, a metacognitive feeling will emerge when there are substantial changes in the expected dynamics, and the particular object will correspond with the cognitive process in question. Imagine that you are star gazing and you fail to name what looks like a familiar constellation — you will probably get a tip-of-the-tongue (ToT) state directed at your astronomical knowledge. If you then try using your fingers to measure angular distance between stars but you misremember the method and things don’t seem to add up, you might get a feeling of error directed at this embodied skill. If you take out your phone and try to use a star-gazing app, and the app does not work, your metacognitive feeling (probably the reverse of confidence) will this time be directed at the app. The above star-gazing sketch provides a good illustration of how metacognitive feelings modulate cognitive processes that distribute across brain, body and world, and of how they elicit disengagement and engagement with cognitive processes at a variety of scales. In terms of AIT, a distributed cognitive process (i.e. the particular object) will have certain associated expected Error dynamics, which will give rise to a formal object that models the dynamics (e.g. categorising them as a ToT state, a feeling of error, etc) and transforms the subject’s affordance landscape (via action policies). This affordance landscape transformation will result in modulation (e.g. guiding the mental search for the name of the constellation, alternating the way we frame stars with our fingers, changing the settings in the phone) and, if unsuccessful, in eventual disengagement (e.g. giving up the mental search and trying to measure the stars with our fingers, or giving up that method and taking out the phone). As we saw in chapter 6, formal objects comprise both action policies and models of cognitive processes, so often a change in action policies will be accompanied with a change in how the expected dynamics are modelled. Imagine a subject is asked for the capital of Serbia. At first, she feels that she knows the answer (feeling of knowing) and tries to recall it. She is unsuccessful and undergoes a ToT state. After a lot of mental effort, she finally gives up in frustration. She will
— 113 —
probably experience not only a change in action policies (retrieval and subsequent disengagement from retrieval), but also a change in the experienced metacognitive feeling (from a feeling of knowing to a ToT state and from that ToT state to the feeling of not knowing). If a formal object provides a model of the dynamics (knowing the answer) and an action policy (retrieval) to regulate those dynamics, the failure of said action policy to regulate the dynamics will be seen as support for a different hypothesis (that the subject does not know the answer), leading to a new formal object both in the form of a new action policy (stopping the retrieval process) and of a new descriptive model of the dynamics (the subject not knowing the answer). The scenario above also shows how metacognitive feelings emerge in succession in order to guide distributed cognitive processes in a dynamic, flexible way. An interesting aspect of these dynamics is how often a subject switches between cognitive processes, rather than gradually disengaging with one process and engaging with a different process. This is particularly important for our purposes because in experimental conditions, disoriented subjects were reported to switch from online to offline systems of spatial representation (Waller and Hodgson 2006). In PP, switches are traditionally conceived of as rapid changes in hypotheses, the standard case study being perceptual switches during binocular rivalry (Hohwy et al. 2008). Of course, hypotheses comprise not only predictions of perceptual stimuli, but also proprioceptive predictions (i.e. actions). A switch between cognitive processes should be then conceived of as a switch between action policies resulting in a rapid rearrangement of the affordance landscape (and possibly a change in the descriptive aspects of the formal object as well). Just to clarify, there is no need to speculate a tight, bidirectional correspondence between distributed cognitive processes and action policies. For the present purposes, it suffices to say that when a metacognitive feeling elicits a switch between cognitive processes it does so via a switch in action policies. An interesting question is which feelings, and under what conditions, will tend to provoke a switch rather than a gradual disengagement. This is primarily an empirical question, but we can advance the following theoretical conjecture: Switches will happen when the organism needs to adapt to what are expected to be fast-changing Error dynamics. In terms of AIT, expected fast-changing Error dynamics correspond to a high Expected Change in Rate, that is, to high levels of arousal. Therefore, the higher its level of arousal, the more likely an affective state with negative valence29 will be to elicit switches between action policies with the corresponding switch between distributed cognitive processes. This idea receives indirect support from the arousal-biased competition theory (Mather and Sutherland 2011), which proposes that when a signal is assigned high priority, then arousal will amplify this signal, leading to a competitive advantage in biased competition (see Lee et al 2014 for a neuroscientific study in support of arousal-biased competition theory). When we put the arousal-biased competition theory in terms of PP and affordance competition, arousal accelerates the selection process between competing hypothesis, resulting in switches rather than gradual transformations (for a study showing the correlation between arousal and switching rate in binocular rivalry experiments see Carter et al 2007). A related finding is that arousal increases learning rate (Nassar et al 2012), which further supports the AIT understanding of arousal, as one would expect the brain to update hypothesis in the light of sensory income at a higher rate the faster the Error dynamics are expected to change. At a fundamental level, affective states contribute to the attunement within a cultural practice. Back to the OAC case, a subtle feeling of error will work as a feedback mechanism to ensure that the player minimises changes in optical acceleration and catches the ball. The goal of catching the ball will result in a prediction of stability in optical acceleration, and a feeling will emerge if there are expected Errors related to that prediction, thus regulating movement to ensure that the dynamics of the player are coupled to those of the ball. What the OAC case reveals is that there is no need for 29 If Rate is expected to increase quickly (i.e. high arousal and positive valence), then one should continue the ongoing
process.
— 114 —
the Errors to climb to high levels of the hierarchy. Once the right predictions have been established high up the prediction hierarchy (e.g. the goal to minimise optical acceleration), a feeling of error will dynamically monitor the situation and guide the organism through the necessary adjustments by subtly tweaking the pattern of precision weightings. If the expectation is of rapid, negative changes in Rate, more resources will be employed to control Error dynamics, and a change in action policy will become more likely. Again, arousal will influence behaviour in a context-sensitive way. In a windy day, high arousal will result in more resources being employed, as long as the ball is perceived to be catchable. And if the ball is going too fast, this will very quickly result in abandoning the OAC strategy, with the corresponding disappearance of the catchability affordance of the ball, as shown in recent experimental work (Postma et al 2018). A final element developed in chapter 7 that we need to cast in AIT terms concerns dual process theories, as the claim was that affective states elicit switches not only between cognitive processes distributed at different scales but also between “fast” and “slow” modes of cognition (M1>M2, M4>M3, etc). In PP, fast and slow modes of cognition are conceptualised within the predictive hierarchy. High level goals contextualise the less flexible responses at lower levels (Pezzulo et al 2015). Whether slow or fast modes of cognition are selected depends on precision weightings along the hierarchy. The larger the precision gain assigned to the higher (lower) levels, the more likely slow (fast) modes of cognition are to be selected. “The result is a continuum between goal-directed and habitual action that depends on the relative weight assigned to the different layers. Habits can arise in this scheme too, when the lower layers acquire sufficient precision to become essentially impermeable to the influences of the higher layers” (p.14, Pezzulo 2017). The link to AIT is fairly straightforward: affective states result in a quick reorganisation of precision weighting (i.e. a transformation in the affordance landscape). This conception of inter-mode switches is congenial with the PP idea that contextually recruited patterns of precision weighting “accomplish a form of set-selection or strategy switching” (p. 14, Clark 2015). Here again, arousal plays an important role. The higher the Expected Change in Rate, the lower the selected response will tend to be along the hierarchy, because lower levels operate in shorter timeframes. This is consistent with much of the experimental work in arousal. For instance, Wichary and colleagues found that negative arousal makes subjects more likely to pursue simpler strategies in decision-making tasks (Wichary et al 2015). Of course, this means that high level of arousal might sometimes lead to lower, faster, more automatic behaviours in non-optimal ways. A recent study illustrates this danger by showing how arousal increases the likelihood of subjects falling prey to fraudulent advertisement (Kircanski 2017). The emerging picture is one of metacognitive feelings —in the form of descriptive-directive models of expected Error dynamics— guiding the agent through fast and slow cognitive processes distributed at various spatiotemporal scales across a rich cognitive ecosystem. Metacognitive feelings contribute to the cohesion of emerging cultural practices through modulation-guided attunement of coupled dynamics, and they contribute to the agent’s disengagement from cultural practices by eliciting a switch between different action policies (or a gradual shift, depending largely on arousal levels). I had argued in chapter 6 that metacognitive feelings evaluate and regulate cognition in order to direct uncertainty to the optimal levels of PEM over time, and we have seen in the previous section that cultural practices tend to reduce entropy over time. Metacognitive feelings also play an important role here, as they guide the subject to engage with those practices that are estimated to be most conducive to PEM over time. Most of these will be negentropic cultural practices that tend to increase short-term predictability. Some of these will be entropic practices of both the learning and the disruptive types. Learning practices will tend to elicit positive valence as the subject becomes better at reducing Error in a new domain. Disruptive practices are again the hardest case to crack. A quick phenomenological reflection points to the feeling of anticipation and of impending insight of such practices, and, following AIT, we might
— 115 —
conjecture that this phenomenology corresponds to a promise of an approaching surge in prediction success, which pushes the subject to engage joyfully with the surrounding uncertainty. This is a topic that we will come back to when we revisit the feeling of manakan in Evenki culture in section 8.5. Before that, let us look at how AIT can develop the characterisation of disorientation as a metacognitive feeling into a full-blown theory of the phenomenon.
III – DISORIENTATION REVISITED Let us take the account of affective experience that emerged in the previous section to cast a light on the claim that disorientation is a metacognitive feeling that evaluates and regulates the subject’s active navigational processes. The particular object of disorientation corresponds to the subject’s active navigational processes. The formal object corresponds to a lack of confidence on said processes. When the active navigational processes are expected to decrease their performance and generate negative Rate (i.e. negative valence), disorientation emerges. The resulting action policies will depend on context and arousal. Much of our discussion in chapters 1-4 revolved around the online/offline and the allocentric/egocentric distinctions concerning spatial representations, which have not been accommodated within the PP framework yet. A good starting point is revisiting the idea of the cognitive map, which I briefly introduced in chapter 1, where I pointed out how the spatial representation system seems to work in a constraint satisfaction mode, with different cell networks in the hippocampal and parahippocampal regions tracking available topological (boundary cells), metric (grid cells), angular (head direction cells) and location (place cells) representations. Despite the astonishing progress in the neuroscience of spatial cognition, there are still many loose ends in our understanding. These include the dependence of place fields on topology and goals (Pfeiffer and Foster, 2013), the heterogeneity of remapping phenomena (Colgin, Moser and Moser 2008; Kubie, Levy, Fenton 2019), and internally generated hippocampal sequences during sleep that go beyond past trajectories (Gupta et al., 2010). A recent trend that is making considerable headway in addressing some of these issues are predictive accounts of hippocampal activity. For instance, Stachenfeld and colleagues propose that place cells encode a predictive representation of future states given the current state, which they show to result in predictive maps that adapt to task structure, thus addressing the issue of place fields being dependent on topology and behaviour policies (Stachenfeld, Botvinick, Gershman, 2017). Another study shows how a framework of the hippocampus as performing hidden state inference can offer a unified theory of remapping (Sanders, Wilson and Gershman, 2020). The most relevant of these predictive accounts for the scope of this chapter is a recent computational theory that views the hippocampus as a generative hierarchical model (Stoianov, Maisto and Pezzulo, 2020). According to this theory, the generative model organizes sequential experiences (trajectories) into coherent spatiotemporal contexts (maps). An important element here is that generative replay (the offline sampling of fictive experiences) gives rise to the internally generated hippocampal sequences, which provides a model for the internal generation of sequences during sleep. The resulting maps are biased, encoding the probability of items at given locations in a way that signals their importance (e.g. goal locations are highly weighted). The generative model is updated in a Bayesian way using current trajectories in order to infer the spatial configuration of the environment. An upshot of this view is that the hippocampal hierarchical generative model “uses both maps and sequences as structures (or inductive biases) to organize items of experience” (p. 15, Stoianov, Maisto and Pezzulo 2020).
— 116 —
The view of the hippocampus as a generative hierarchical model is relevant here because it fits seamlessly within the PP framework30, and because it provides a solid link between spatial representation and spatial experience. The study by Stoianov, Maisto and Pezzulo offers an empirically consistent computational account of spatial representation, one in which predictions at a high level in the hierarchy (maps) influence predictions at lower levels (sequences and items of experience). For matters of simplicity, I will refer to the level of the hierarchy producing maps as L2 and to the level of the hierarchy corresponding to items of experience as L1. This view joins the PP aspects of this monograph with the online-offline spatial representation distinction introduced in chapter 1. An offline spatial representation is a decoupled generative model. In turn, the online spatial representation is the spatial configuration of the environment inferred given previous experience and current input. In other words, it is the most likely hypothesis regarding spatial configuration, and, consequently, the one populating consciousness. The update of the generative model corresponds to the integration of allocentric (maps of the environment at high levels of the hierarchy) and egocentric (the spatial configuration of nearby items around the subject) spatial frames of reference. Following this view, we finally have a connection between a cognitive map and the horizon of intentionality that structures our experience of the environment. Disorientation emerges when there is an expectation of negative Rate that is inferred to be caused by a deficiency in the active navigational processes. Let us take a standard case of disorientation, such a subject losing their way in a mall (e.g. the subject of scenario 1 back in chapter 4). Inaccurate L2 predictions result in increasing Error at lower levels (e.g. there is a prediction that after turning a corner, a corridor will appear on the left, but the corridor appears on the right), the system infers that there is a problem in the active navigational processes and disorientation emerges. If a subject is lost, there is good reason to expect a cascading effect of increasing Error, which means that a mismatch between L2 and current Error might lead to a relatively sharp (i.e. highly arousing) and strong (i.e. high intensity) episode of disorientation, as the system expects a surge in forthcoming Error. Of course, the cause of the expected increase in Error will tend to be context-dependent and more specific than a deficiency in active navigational processes, which is a fairly broad umbrella term. If the system infers that the issue is due to an inaccurate cognitive map, the predictions at L2 will update to accommodate the incoming error, which will cause a switch in the overall hypothesis, and the individual will start to explore and learn the environment anew (forming new predictions at L1 and L2). On the other hand, if the system infers the issue to be a problem in allocentric-egocentric integration (i.e. a problem in exploiting the predictions at L2 to generate accurate predictions at L1), there will be a sustained effort in re-interpreting the surroundings to figure out how they fit with the cognitive map. This is the case of many turned-around episodes. The individual believes to be facing Place de la République (the result of the integration of L2 and L1), realizes to be facing Hotel de Ville, and after a surge of disorientation (the disintegration of L2 and L1) the environment seems to rotate 180º (a reintegration of L2 and L1). There is good reason, however, to refer to active navigational processes broadly, rather than to narrow down on a single navigational process (or a set of them), such as allocentric-egocentric integration. Here again, think of following the instructions of a GPS device that suddenly starts to glitch. This will cause an expectation of increase in Error, the inferred cause of which will be the active navigational processes. Only in this case, the issue is not the accuracy of internally generated maps at L2, or of the integration of these maps with lower-level predictions and Error; the issue is the malfunctioning of a cognitive artefact. A borderline case might be a map, because it is hard to tell if the hippocampus would generate new high-level models based on the map (at L2) or if the map would directly be the source for generating the predictions at L1, the latter being more 30 The article in question is the first research output of a large ERC project (ThinkAhead) led by Pezzulo to
conceptualise human planning (navigation being a chief topic) within the PP framework.
— 117 —
parsimonious and streamlined. At the moment, this is an open question, which might eventually be solved through empirical research. Although external representations is not an issue that is directly addressed by Stoianov, Maisto and Pezzulo, their system accommodates the coordination of the hippocampus and the mPFC for bi-directional learning: “cortical systems can transfer structured, rule-related knowledge to hippocampal systems, potentially contributing to bias their content and reorganize experiences according to learned rules . A theoretical possibility that remains to be tested in future research is whether during periods of coupled oscillations, hippocampal and cortical structures simultaneously replay from their respective generative models and train each other.” (p. 14, Stoianov, Maisto and Pezzulo, 2020). This theoretical possibility opens the door to cortically mediated hippocampal learning of external representations. Still, as the GPS guidance scenario highlights, disorientation will also emerge when the active navigational processes are not primarily reliant on hippocampal models (e.g. as is the case with what we deemed M4 navigation in chapter 7). The same applies to following a guide, or a signage system. The view of the hippocampus as a generative hierarchical model provides an elegant and PP-congenial way to conceptualise the standard case of disorientation, but it is important to broaden our scope, because disorientation evaluates navigation in a flexible way that goes beyond hippocampally generated representations of space. As we saw in the previous section, arousal will play an important role in determining whether the emergence of a metacognitive feeling (in this case, disorientation) results in modulation of cognitive processes (the active navigational processes) or a switch (away from the current cognitive processes). Ceteris paribus, high-arousal situations will result in ‘fast’ responses (e.g. exploring, random walk…) as opposed to ‘slow’ responses (e.g. deliberation). The higher (lower) the arousal (i.e. Expected Change in Rate), the larger the precision gain associated to lower (higher) levels and the more (less) likely that the emergence of disorientation will result in a switch. Low arousal episodes of disorientation will tend to lead to explicit deliberation and to modulation of active cognitive processes. High arousal episodes of disorientation will tend to lead to faster, lower-level action responses and to a rapid collapse of the current overall hypothesis in order to find a new model that best accommodates the lower level input. Because of the directive-descriptive nature of generative models, the collapse will take the form of a switch of both action policy and of the inferred spatial configuration of the environment. For instance, when a subject is ‘turned around’, a new integrated model of the environment will emerge, in which the egocentric-allocentric integration shifts 180º. And when a subject loses their way out of the subway, they might take out their GPS-device and engage in an alternative distributed cognitive process. However, if there is no alternative viable model of the environment, no new stable inferred spatial configuration will emerge. This will result in exploratory behaviour (and often in suboptimal behaviour such as random walking) and in the transformation of the subject’s experience that we first explored in chapter 2.
IV – THE PHENOMENOLOGY OF DISORIENTATION REVISITED Let us now establish how negative Expected Rate inferred to be caused by a deficiency in the active navigational processes results in the subject undergoing a reduction of their possibility space within an unfamiliar environment framed by an uncertain horizon of experience. In other words, let us connect the PP account of disorientation with the phenomenological account of disorientation presented in chapter 2. Although not the standard case of disorientation, Binet’s turned-around subjects provide again a go-to case to start the discussion. We had highlighted the following aspects in the experience of being turned around:
— 118 —
• egocentric frames of reference are not affected by it; • the allocentric frames of reference are almost always misaligned 180º; • it often prevents recognition of well-known environments;
At this point, it should be clear that what goes awry in the turned-around scenario is the integration of allocentric and egocentric frames of reference. Although L2-L1 integration is not discussed in Stoianov, Maisto and Pezzulo’s account, it seems clear from the standard PP view that this integration is the result of a dynamic process of Error updating, so that as the subject moves, bringing new items into view, the predictions generated at L2 adjust accordingly. And again, maps at L2 are structures (qua inductive biases) that organise items of experience. In the turned-around scenario, even though the low-level predictions regarding the spatial relation of the agent to the proximal environment (i.e. minimal self-location) are accurate, the predictions regarding the spatial relation of the agent to the non-proximal environment are inaccurate. As I argued in chapter 1, the common 180º misalignment results from the role of the three bodily axes in the allocentric-egocentric integration. In PP terms, the subject’s body is used to infer the position of the agent with respect to the generative models produced at L2. In the most common case, the subject turns left instead of right by mistake (or vice versa) and this results in a 180º misalignment of the spatial items that are expected to be encountered on the way. If the body serves to infer L2-L1 integration, this integration in turn serves to infer the nature of the encountered items in the subject’s experience, which often results in the failure to recognise well-known environments for turned-around subjects. Our much-quoted turned-around subject believing to be going towards Place de la République takes a while to recognise Hotel de Ville, because it was unexpected. This points to the role of L2 maps in anticipating the identity of the items encountered at L1. This anticipation is an important element of the structural role of the horizon of experience. In chapter 2, I conceptualised the analogy between orientation illusions and figure-ground illusions (first introduced by Charles Olman in the context of microgravity environments) in terms of a transmodal instability in egocentric-allocentric integration that is similar to the perceptual instability in figure-ground illusions. The elaborate picture that is emerging in this chapter gives us a much more detailed story. Here, the instability of disorientation corresponds to the decrease in confidence about the L2 generative model. Think back to the PP explanation of binocular rivalry (introduced in chapter 5): the most likely explanation of the visual input switches between the house and the face hypothesis, none of which is stable (as neither can explain away the Error coming from one of the eyes). Sometimes, there are also brief and untenable face-house-blur moments right before a switch. This is similar to the turned-around scenarios: The Hôtel de Ville appears unexpectedly, which causes Error to surge. Two things happen at once: an expectation of incoming navigation-related Error (leading to disorientation) and the current L2 model becoming unable to explain away Error. The whole spatial configuration of the environment seems to hover around the subject for a moment (as the current L2 model becomes fails to explain away Error), and, upon recognising the building, the environment seems to turn 180º around the subject (as a new L2 model emerges). Slowly, the feeling of disorientation dissolves, because as the new model explains away incoming Error, the expectation of negative Rate dissipates. Now, we can also explain why the instability in question is transmodal: the relevant model is at a high level and only as predictions cascade down the hierarchy are modality-specific predictions generated. In other words, the L2 generative model is not modality specific. Most of what we have said about being turned around can be applied to the standard case of disorientation (i.e. to the instability in allocentric-egocentric integration): the model at L2 fails to explain away Error and this results in transmodal instability and in Expected negative Rate attributed to the active navigational processes; and when a new model explains away the Error, the expectation of negative Rate dissipates. In PP-neutral terms, there is a spike of uncertainty about the spatial configuration of the
— 119 —
environment in relation to the subject and this results in disorientation — a lack of confidence about the active navigational processes. In chapter two, I broached the anxiety, helplessness, confusion and diminishment associated with disorientation, and grouped these affective aspects of the experience under the umbrella of body-space reduction, characterised by a diminished sense of possibility. This diminished sense of possibility is best explained as an impoverishment of the subject’s affordance landscape. Remember how our turned-around subject knows that the building in front of him (Hôtel the Ville) can be entered, but not that he himself can enter it. The building is no longer familiar, and he does not feel welcomed to explore it. There is a plethora of possibilities for action that depend on successful navigation, ranging from the ability to enter familiar buildings to our ability to find our way back home. The emergence of the metacognitive feeling of disorientation transforms the affordance landscape by assigning low precision to these navigation-dependent affordances. This, experientially intertwined with negative valence (and possibly high arousal), results in the afore-mentioned anxiety, helplessness, confusion and diminishment that are characteristic of disorientation.
V – BACK TO MANAKAN At first glance, it might seem that AIT is ill-suited to account for the positively-valenced episodes of disorientation explored in Chapter 3. After all, positive valence corresponds to positive Expected Rate, and it is hard to see how a failure in the active navigational processes could result in an expectation of a decrease in Error. The failure of a predictive model seems at odds with the expectation of an increase in prediction success. This case of ‘joyful disorientation’ is the last (and possibly the toughest) nut to crack for the present account. As we did in chapter 3, we will start by looking at the experience of manakan among the Evenki, which will serve to illuminate the functional dynamics of the ‘joyful disorientation’ case. In chapter 3, we had clarified that the unprescribed experience of the environment during manakan is only made possible because for the Evenki there is no ongoing integration of egocentric and allocentric frames of reference during wayfinding, given that the allocentric-egocentric integration overdetermines the subject’s horizon of experience through expectation and anticipation. In contrast, a complex navigational system (involving path networks, and vast hydrological and toponymical knowledge) allows the Evenki to attain a skilful attunement to the environment and to open themselves to its possibilities. We have seen that predictions at L2 (maps) constraint predictions at L1, and the result of this in the subject’s horizon of experience. The degree to which L2 will constraint L1 depends on precision weighting. In a paradigmatic case of survey-knowledge navigation, the models at L2 are assigned high precision with respect to L1, so that Error that does not adjust to L2 will be more likely to be ignored. For individuals relying on practical mastery such as the Evenki, the models at L2 are assigned low precision, so that Error will tend to be resolved at lower levels of the hierarchy without relying on L2 models31. In other words, cognitive maps will not play a substantial role in structuring the subject’s environmental experience. As a result, the spatial configuration of the
31 A question that is outside the scope of this chapter is the degree to which those L2 models are even generated in the
case of the Evenki, or if they are simply non-existent. It might be that such a nesting of generative models depends on culture-specific developmental processes. On the other hand, because Evenki rely on route knowledge, it seems certain is that some sequences (i.e., paths rather than maps) do play a certain role in constraining L1 predictions. However, it is likely that this results in paths being experienced as affordances to go to certain places and that there is no overarching model of how these sequences fit together.
— 120 —
distal environment will be underdetermined for ‘practical mastery’ individuals compared to ‘mental map’ individuals. The upshot is that Errors arising at lower levels of the hierarchy are resolved by quickly adapting predictions at those lower levels. This is what results in the skilful attunement of the Evenki — it is a quick adaptation to the vicissitudes of a changing environment. The Evenki approach is certainly a model-light, bag-of-tricks approach. It relies on a variety of heuristics and on practical, embodied knowledge, not on a representation of the distal environment. As they advance, the Evenki encounter new places and with them, new hitherto unexpected possibilities for action. Sensing these possibilities is very much an embodied affair, which involves light walking, thin footwear through which one can feel the ground, breaking branches to leave a trail, etc (for a detailed account, see Safonova and Sántha 2013). This is similar to the case of using the AOC strategy to catch the ball, in which a certain embodied practice serves to exploit heuristics without a need for a detailed model of the ball’s trajectory. As with AOC, Evenki walking relies on a flexible precision-weighting pattern to quickly adapt to low level Error without a need for complex models high up in the hierarchy. Lower levels operate at smaller spatial and temporal scales. The spatial aspect means that they involve predictions only about the proximal environment. The temporal aspect means that they involve much thinner anticipation. Low level predictions are more immediate, both spatially and temporally. Consequentially, ‘practical mastery’ navigation results in a more immediate experience, without the degree of expectation and anticipation involved in ‘mental map’ navigation. Practical mastery is a way of embracing uncertainty. It is easy to see how this style of wayfinding results in the sense of possibility and of excitement experienced during Manakan: skilful walking results in a continuous unrolling of new affordances. We can also see how this results in positive valence in the terms of AIT: The ongoing fulfilment of budding affordances and the corresponding expectation of new affordances-to-be-fulfilled results in an expectation of accurate predictions (as affordances result in actions that fulfil proprioceptive predictions) — in other words, the burgeoning affordance landscape corresponds to an expectation of positive Rate. Let us now look at the joyful instances of disorientation. In chapter 3, we had reviewed the existing literature on the subject and listed a series of features associated with such episodes:
• a lack of origin, location and destination, • having no set purpose, • getting beyond simplistic models, • suspending usual ways of movement, • getting drawn by the attractions of the terrain, • the undecidable aspect of spatial experience, • establishing a new kind of relationship with space, • an experience of freedom.
All of the above resonate with Evenki skillful walking and the experience of Manakan as we have just characterized it, except for two features: suspending usual ways of movement and establishing a new kind of relationship with space. Moreover, as we saw in chapter 3, we cannot say that the Evenki are disoriented when they experience Manakan. This is precisely because for the Evenki, skillful walking is their usual way of movement, while the same cannot be said for, say, the Situationists that undertake a dérive. The usual way of movement of most of the subjects undergoing episodes of joyful disorientation is largely based on survey knowledge. Disorientation sometimes results in these subjects adopting a new way of movement, one akin to skillful walking, and as a result they experience a feeling akin to Manakan. This shift is precisely what practices such as the dérive aim to achieve. The dérive is a clear entropic cultural practice: in the short-term, it
— 121 —
increases entropy (disorder, unpredictability) with the aim of establishing alternative cultural practices, i.e. new kinds of relationships with space. Joyful episodes of disorientation then occur when the disorientation results in suspending usual ways of movement and establishing a new kind of relationship with space. In AIT terms, disorientation results in a rearrangement of the pattern of precision weightings: from a model-heavy mode to a mode in which Error is resolved at lower levels of the hierarchy without relying on L2 models. This is what is meant by “getting beyond simplistic models” (simplistic as in schematic, the models can be incredibly complex!), which results in embracing “the undecidable aspect of spatial experience” (as the distal environment is not represented and remains underdetermined) and in “getting drawn by the attractions of the terrain” (the generation of new affordances thanks to this more flexible mode of navigation). This is the experience of freedom that comes in the joyful episodes of disorientation. The process through which joyful episodes of disorientation result in positive valence is analogous to the process through which skilful walking results in positive valence for the Evenki: new possibilities for action emerge and are then fulfilled through action, which results in an expectation of positive Rate. There is a sense then, in which we cannot say that the feeling of disorientation has positive valence: what results in positive valence is not, strictly speaking, the shortcoming of an active navigational process, but the emergence of an alternative navigational process. The shortcoming of an active navigational process will necessarily result in expected negative Rate, but this negative valence will be subsumed into the larger expectation of positive Rate if a new, promising navigational process appears. Hence, the joyful scenarios of disorientation will be a hybrid affair, valence-wise. Disorientation will initially result in negative valence, but as an alternative navigational process appears, the polarity will slowly change. At times, this might even result in mixed feelings, with Rate expected to have different polarities at different timescales (e.g. negative in the initial period of disorientation and positive afterwards), making episodes of positive disorientation akin to walking into a surprise party, savouring a guilty pleasure or remembering bitter-sweet memories. There is always a threat in losing’s one way and for the negative valence of disorientation to be overcome by the positive valence of ‘joyful’ disorientation, the threat of losing one’s way needs to be minor. This is well captured by Ray Bradbury’s dictum: “To be lost. How frightening. To be safely lost. How wonderful” (p. 61, Bradbury 2017). An additional aspect to consider here is that to ever enjoy having “no set purpose”, and “a lack of origin, location and destination”, the subject really needs to have nowhere to get to and no business to attend. Something that becomes clear from the corpus of reports that we have been collecting is that the anxiety of disoriented subjects is strongly connected to both perceived danger
I was walking on narrowed streets with little light and a few people by myself at night. I felt unsafe and anxious because no one was with me, and the environment was not familiar to me.
and to having a destination one is trying to get to I was lost in the NYUAD Arts Center trying to find my classroom. It was my first day of class, and there were minimum signage indicating where the classrooms were. I was anxious really anxious because I didn't want to be late to my first class.
Both of these factors are in line with our account. If the situation is perceived as dangerous, there will be an expectation of negative Rate. If the subject has an active goal (e.g. getting to class), the failure to fulfil that goal will equally result in a spike in Error, and the episode of disorientation will result in an expectation of negative Rate.
— 122 —
VI – CONCLUSION At the heart of this chapter, of this monograph and of the proposed theory is the claim that disorientation is a metacognitive feeling that evaluates and regulates the subject’s active navigational processes. When the active navigational processes are expected to decrease their performance, disorientation emerges. The performance of the active navigational processes (the particular object of the feeling) is then modelled through a formal object (which corresponds to diminishing confidence in said processes) that is both a descriptive model and a directive action policy. The actions resulting from disorientation will depend both on context and on the levels of valence and arousal. Low arousal episodes of disorientation will tend to lead to explicit deliberation and to modulation of active cognitive processes. High arousal episodes of disorientation will tend to lead to faster action responses and to a rapid collapse of the current model and action policy guiding navigation, resulting in a switch between distributed cognitive processes (e.g., from sauntering to active wayfinding, or from following a map religiously to exploring the immediate environment). Regarding the phenomenology of disorientation, two aspects of the theory come to the fore. First, disorientation weakens the influence of spatial models on shaping environmental experience, so that the horizon of intentionality of disoriented subjects becomes less determined as a result of disorientation. Second, in paradigmatic, negative cases, the emergence of disorientation transforms the subject’s affordance landscape by draining (through precision-weighting) navigation-dependent affordances, leading to the anxiety, helplessness, confusion and diminishment that are characteristic of disorientation. Finally, there are some cases in which this re-arrangement of the affordance landscape results in the emergence of alternative navigational processes that bring about the positive aspects of disorientation. Something interesting about the joyful disorientation cases and the skilful walking of the Evenki is that our present considerations throw light on broader issues. The idea that the fundamental activity of the brain is the generation and updating of predictions does not necessarily entail a model-heavy view of cognition. The dexterity of the Evenki at wayfinding shows that cognition (in this case, navigation) can operate at lower, model-light levels and still compare favourably with higher model-heavy cognitive processes. Moreover, the contrast between ‘practical mastery’ navigation and ‘mental map’ navigation shows how our models of the world constraint our experience of that world, and how alternative modes of cognition result in alternative modes of experience. Sometimes, only after our maps have fallen apart does the environment show itself to be ripe with possibility. Here lies the ambivalence of disorientation. To lose our grasp of the world is a dangerous affair, and our feelings are wise to tell us so. And yet, without the momentary shedding of the old ways, new ways would never spring forth.
— 123 —
B I B L I O G R A P H I E
A. S. Gupta, M. A. A. van der Meer, D. S. Touretzky, and A. D. Redish, ‘Hippocampal
Adams, Annette L., et al. “Search Is a Time-Critical Event: When Search and Rescue Missions May Become Futile.” Wilderness & Environmental Medicine, vol. 18, no. 2, June 2007, pp. 95–101. ScienceDirect, doi:10.1580/06-WEME-OR-035R1.1.
Aguirre, G. K., & D’Esposito, M. (1999). Topographical disorientation: a synthesis and taxonomy. Brain, 122(9), 1613–1628.
Allen, G. L. (1999). Spatial abilities, cognitive maps, and wayfinding. Wayfinding Behavior: Cognitive Mapping and Other Spatial Processes, 4680.
Allen, G. L., & Haun, D. B. (2004). Proximity and precision in spatial memory. Human spatial memory: Remembering where, 41-63.
Alsmith, A. J. Perspectival Structure and Agentive Self-Location (2017). In de Vignemont, F., & Alsmith, A. J. (Eds.). The Subject's Matter: Self-Consciousness and the Body (pp. 263-289). MIT Press.
Amorim, M. A., Glasauer, S., Corpinot, K., & Berthoz, A. (1997). Updating an object’s orientation and location during nonvisual navigation: A comparison between two processing modes. Perception & psychophysics, 59(3), 404-418.
Appleyard, D. (1970). Styles and methods of structuring a city. Environment and Behavior, 2(1), 100–117.
Arango-Muñoz, S. (2013). Scaffolded Memory and Metacognitive Feelings. Review of Philosophy and Psychology, 4(1), 135–152. https://doi.org/10.1007/s13164-012-0124-1
Aristotle (1986). De Anima (On the Soul) (Trans. Hugh Lawson-Tancred). Harmondsworth: Penguin Books.
Auvray, M., Myin, E., and Spence, C. (2010). “The Sensory-Discriminative and AffectiveMotivational Aspects of Pain”. Neuroscience and Biobehavioral Reviews, p. 11.
Bain, D. (2013). “What Makes Pains Unpleasant?” Philosophical Studies 166.1, pp. 69– 89.
Bain, D. (2014). Pains that don't hurt. Australasian Journal of Philosophy, 92(2), 305-320.
Barrett, L. F. (2006). “Valence Is a Basic Building Block of Emotional Life”. Journal of Research in Personality 40.1, pp. 35–55.
Barrett, L. F. (2006). Are Emotions Natural Kinds? Perspectives on Psychological Science, 1(1), 28–58. https://doi.org/10.1111/j.1745-6916.2006.00003.x
Barrett, L. F. (2014). The conceptual act theory: A précis. Emotion review, 6(4), 292-297.
Barrett, L. F. (2017). The theory of constructed emotion: an active inference account of interoception and categorization. Social cognitive and affective neuroscience, 12(1), 1-23.
Barrett, L. F. and Bliss-Moreau, E. (2009). “Affect as a Psychological Primitive”. Advances in Experimental Social Psychology. Vol. 41, pp. 167–218.
Barry, Caswell, et al. “The Boundary Vector Cell Model of Place Cell Firing and Spatial Memory.” Reviews in the Neurosciences, vol. 17, no. 1–2, 2006, pp. 71–97.
Batson, C. D., Shaw, L. L., and Oleson, K. C. (1992). “Differentiating Affect, Mood, and Emotion: Toward Functionally Based Conceptual Distinctions”. Emotion, pp. 294– 326.
Beran, M. J. (2012). Foundations of Metacognition. OUP Oxford.
Bermúdez, J. L. (1998). The paradox of self-consciousness. MIT Press
Bermúdez, J. L. (2017). Understanding" I": Language and Thought. Oxford University Press.
Berridge, K. C. and Kringelbach, M. L. (2015). “Pleasure Systems in the Brain”. Neuron 86.3, pp. 646–664.
Berthier, M., Starkstein, S., and Leiguarda, R. (1988). “Asymbolia for Pain: A SensoryLimbic Disconnection Syndrome”. Annals of Neurology 24.1, pp. 41–49.
Best, G. (1970) Direction Finding in Large Buildings, in D. V. Canter (ed.), Architectural Psychology, pp. 72-75, London, RIBA
Binet, M. A. (1885). Vertigo of Direction. Mind, 10 (8), 156-159
Binet, M. A. (1894). Reverse illusions of orientation. Psychological Review, 1(4), 337
Blakemore, R. L., & Vuilleumier, P. (2017). An emotional call to action: Integrating affective neuroscience in models of motor control. Emotion Review, 9(4), 299-309.
Blanke, O., & Metzinger, T. (2009). Full-body illusions and minimal phenomenal selfhood. Trends in cognitive sciences, 13(1), 7-13.
Blodgett, H. C. “The Effect of the Introduction of Reward upon the Maze Performance of Rats.” University of California Publications in Psychology, vol. 4, 1929, pp. 113–34.
Boldt, A., & Gilbert, S. J. (2019). Confidence guides spontaneous cognitive offloading. Cognitive Research: Principles and Implications, 4(1), 45. https://doi.org/10.1186/s41235-019-0195-y
Bradbury, R. (2017). Yestermorrow: Obvious answers to impossible futures. Rosetta Books.
Brandišauskas, D. (2009). Leaving footprints in the Taiga: Enacted and emplaced power and luck among the Orochen-Evenki of the Zabaikal Region in East Siberia. University of Aberdeen.
Brandišauskas, D. (2012). Making a Home in the Taiga: Movements, Paths and Signs among Orochen-Evenki Hunters and Herders of Zabaikal Krai (South East Siberia). Journal of Ethnology and Folkloristics, 6(1), 9–25.
Brewer, B. (1992). Self-location and agency. Mind, 101(401), 17-34.
Brown, C. A., Seymour, B., Boyle, Y., El-Deredy, W., and Jones, A. K. (2008). “Modulation of Pain Ratings by Expectation and Uncertainty: Behavioral Characteristics and Anticipatory Neural Correlates”. Pain 135.3, pp. 240–250.
Brown, R., & McNeill, D. (1966). The “tip of the tongue” phenomenon. Journal of verbal learning and verbal behavior, 5(4), 325-337.
Bufacchi, R. J., & Iannetti, G. D. (2016). Gravitational cues modulate the shape of defensive peripersonal space. Current Biology, 26(21), R1133-R1134.
Bufacchi, R. J., & Iannetti, G. D. (2018). An action field theory of peripersonal space. Trends in cognitive sciences. 22(12), 1076.
Burgess, N., Trinkler, I., King, J., Kennedy, A., & Cipolotti, L. (2006). Impaired allocentric spatial memory underlying topographical disorientation. Reviews in the Neurosciences, 17(1-2), 239-252.
Byrne, Patrick, et al. “Remembering the Past and Imagining the Future: A Neural Model of Spatial Memory and Imagery.” Psychological Review, vol. 114, no. 2, 2007, pp. 340–75. APA PsycNET, doi:10.1037/0033-295X.114.2.340.
Campbell, C. A. (2001). The Evenki system of paths: The study of travel and technology in east-central Siberia.
Campbell, J. (1995). Past, space, and self. MIT Press.
Carl Georg Lange. (1887). Ueber Gemüthsbewegungen eine psychophysiologische Studie. T. Thomas. Retrieved from http://archive.org/details/bub_gb_az8BgMIeTpoC
Carruthers, P. (2017). “Valence and Value”. Philosophy and Phenomenological Research 97.3, pp. 658–680.
Carver, C. (2003). Pleasure as a sign you can attend to something else: Placing positive feelings within a general model of affect. Cognition and Emotion, 17(2), 241–261. https://doi.org/10.1080/02699930302294
Carver, C. S., & Scheier, M. F. (1990). Origins and functions of positive and negative affect: A control-process view. Psychological Review, 97(1), 19–35. https://doi.org/10.1037/0033-295X.97.1.19
Carver, C. S., & Scheier, M. F. (2001). On the Self-Regulation of Behavior. Cambridge University Press.
Casati, R. (2017). Two, Then Four Modes of Functioning of the Mind: Towards a Unification of “Dual” Theories of Reasoning and Theories of Cognitive Artifacts. In Representations in Mind and World. Routledge. https://www.taylorfrancis.com/
Cassam, Q. (1997). Self and world. Oxford: Oxford University Press.
Castro, B. O. D., Veerman, J. W., Koops, W., Bosch, J. D., & Monshouwer, H. J. (n.d.). Hostile Attribution of Intent and Aggressive Behavior: A Meta-Analysis. Child Development, 73(3), 916–934. https://doi.org/10.1111/1467-8624.00447
Charland, L. (2005). “The Heat of Emotion: Valence and the Demarcation Problem”. Journal of Consciousness Studies 12.8-10, pp. 8–10.
Charland, L. (2005). “The Heat of Emotion: Valence and the Demarcation Problem”. Journal of Consciousness Studies 12.8-10, pp. 8–10.
Cherubini, A., & Chaumette, F. (2010). A redundancy-based approach for obstacle avoidance in mobile robot navigation. In 2010 IEEE/RSJ International Conference on Intelligent Robots and Systems (pp. 5700-5705). IEEE.
Cheung, A., Ball, D., Milford, M., Wyeth, G., & Wiles, J. (2012). Maintaining a cognitive map in darkness: the need to fuse boundary knowledge with path integration. PLOS Computational Biology, 8(8), e1002651.
Chua, E. F., & Bliss-Moreau, E. (2016). Knowing your heart and your mind: The relationships between metamemory and interoception. Consciousness and Cognition, Consciousness and Cognition, 45, 45, 146, 146–158.
Cisek, P. (2007). “Cortical Mechanisms of Action Selection: The Affordance Competition Hypothesis”. Philosophical Transactions of the Royal Society B: Biological Sciences 362.1485, pp. 1585–1599.
Clark, A. (2013). “Whatever next? Predictive Brains, Situated Agents, and the Future of Cognitive Science”. Behavioral and brain sciences 36.3, pp. 181–204.
Clark, A. (2015). Surfing uncertainty: Prediction, action, and the embodied mind. Oxford University Press.
Clore, G. L. (1994). “Why Emotions Are Never Unconscious”. The Nature of Emotion: Fundamental Questions, pp. 285–290.
Colgin LL, Moser EI, Moser MB. 2008. Understanding memory through hippocampal remapping. Trends in Neurosciences 31:469–477. DOI: https://doi.org/10.1016/j.tins.2008.06.008, PMID: 18687478
Colgin, Laura Lee, et al. “Understanding Memory through Hippocampal Remapping.” Trends in Neurosciences, vol. 31, no. 9, Sept. 2008, pp. 469–77. ScienceDirect, doi:10.1016/j.tins.2008.06.008.
Colombetti, G. (2005). “Appraising Valence”. Journal of Consciousness Studies 12.8-10, pp. 8–10.
Colombetti, G. and Harrison, N. (2018). From Physiology to Experience: Enriching Existing Conceptions of “Arousal” in Affective Science. In The Interoceptive Mind: From Homeostasis to Awareness, pp. 245–258
Cooper, H. S. (1976). A house in space. Holt, Rinehhart and Winston, Austin, TX.
Corder, G., Ahanonu, B., Grewe, B. F., Wang, D., Schnitzer, M. J., and Scherrer, G. (2019). “An Amygdalar Neural Ensemble That Encodes the Unpleasantness of Pain”, p. 7.
Corns, J. (2014). “Unpleasantness, Motivational Oomph, and Painfulness”. Mind & Language 29.2, pp. 238–254.
Craig, A. D. (2003). “Interoception: The Sense of the Physiological Condition of the Body”. Current Opinion in Neurobiology 13.4, pp. 500–505.
Craig, A. D. (2009). How do you feel--now? The anterior insula and human awareness. Nature Reviews. Neuroscience, 10(1), 59–70. https://doi.org/10.1038/nrn2555
Cramer, K. (1970). Erlebnis. Thesen zu Hegels Theorie des Selbstbewußtseins mit Rücksicht auf die Aporien eines Grundbegriffs nachhegelscher Philosophie. Stuttgarter Hegel-Tage, 537-603.
Critchley, H. D., Wiens, S., Rotshtein, P., Öhman, A., & Dolan, R. J. (2004). Neural systems supporting interoceptive awareness. Nature Neuroscience, 7(2), 189–195. https://doi.org/10.1038/nn1176
Critchley, H. D., Wiens, S., Rotshtein, P., Ohman, A., and Dolan, R. J. (2004). “Neural Systems Supporting Interoceptive Awareness”. Nature Neuroscience 7.2, p. 189.
Csikszentmihalyi, M. (2009). Creativity: Flow and the Psychology of Discovery and. Harper Collins.
Czaplicka, M. A. (1916). My Siberian Year. Mills & Boon.
Damasio, A. and Carvalho, G. B. (2013). “The Nature of Feelings: Evolutionary and Neurobiological Origins”. Nature Reviews Neuroscience 14.2, pp. 143–152.
Damasio, A. R. (1994). Descartes’ Error: Emotion, Rationality and the Human Brain.
Damasio, A. R. (1999). The Feeling of what Happens: Body and Emotion in the Making of Consciousness. Harcourt Incorporated.
Davies, M. (2000) Interaction without reduction: The relationship between personal and sub-personal levels of description. Mind & Society 1, 87–105
Davydov, V. (2013). Vladimir Davydov, People on the Move: Development Projects and the Use of Space by Northern Baikal Reindeer Herders, Hunters and Fishermen. Études Mongoles et Sibériennes, Centrasiatiques et Tibétaines, 43–44.
Dawson, M. E., Filion, D. L., & Schell, A. M. (n.d.). Is Elicitation of the Autonomic Orienting Response Associated With Allocation of Processing Resources? Psychophysiology, 26(5), 560–572. https://doi.org/10.1111/j.1469-8986.1989.tb00710.x
De Haan, S., Rietveld, E., Stokhof, M., and Denys, D. (2013). “The Phenomenology of Deep Brain Stimulation-Induced Changes in OCD: An Enactive Affordance-Based Model”. Frontiers in Human Neuroscience 7.
De Monticelli, R. (1986). On orientation. Topoi, 5(2), 177-185.
De Sousa, R. (1987). The Rationality of Emotion. MIT Press.
Debord, G. (2012). Theory of the Derive.”(1958). Situationist International Anthology, 50–54.
Dennett, D.C. (1969) Content and Consciousness. London: Routledge and Kegan Paul.
Deonna, J. A. and Teroni, F. (2017). “Getting Bodily Feelings Into Emotional Experience in the Right Way”. Emotion Review 9.1, pp. 55–63.
Desantis, A., Hughes, G., & Waszak, F. (2012). Intentional Binding Is Driven by the Mere Presence of an Action and Not by Motor Prediction. PLOS ONE, 7(1), e29557. https://doi.org/10.1371/journal.pone.0029557
Dhar, M., Wiersema, J. R., & Pourtois, G. (2011). Cascade of neural events leading from error commission to subsequent awareness revealed using EEG source imaging. PLoS One, 6(5), e19578
Doherty, Paul J., et al. “An Analysis of Probability of Area Techniques for Missing Persons in Yosemite National Park.” Applied Geography, vol. 47, Feb. 2014, pp. 99–110. ScienceDirect, doi:10.1016/j.apgeog.2013.11.001.
doi: 10.1016/j.neuron.2010.01.034.
Dokic, Jérôme. Seeds of self-knowledge: noetic feelings and metacognition. In Beran, M. J., Perner, J., & Proust, J. (2012). Foundations of Metacognition. OUP Oxford.
Döring, S. A. (2007). “Seeing What to Do: Affective Perception and Rational Motivation”. dialectica 61.3, pp. 363–394.
Dreyfus, H. L. (1991). Being-in-the-world: A commentary on Heidegger's Being and Time, Division I (Vol. 1). Mit Press.
Dudchenko, P. A. (2010). Why People Get Lost: The Psychology and Neuroscience of Spatial Cognition. Oxford University Press.
Dumont, A. (2018). Are the Evenki reindeer herders still nomads? The alternate use of different types of spaces in Inner Mongolia, China. Études Mongoles et Sibériennes, Centrasiatiques et Tibétaines, 49.
Duncan, S. and Barrett, L. F. (2007). “Affect Is a Form of Cognition: A Neurobiological Analysis”. Cognition & Emotion 21.6, pp. 1184–1211.
Endrass, T., Franke, C., & Kathmann, N. (2005). Error awareness in a saccade countermanding task. Journal of Psychophysiology, 19(4), 275-280.
Endrass, T., Reuter, B., & Kathmann, N. (2007). ERP correlates of conscious error recognition: aware and unaware errors in an antisaccade task. European Journal of Neuroscience, 26(6), 1714-1720.
Epstein, R. A., Patai, E. Z., Julian, J. B., & Spiers, H. J. (2017). The cognitive map in humans: spatial navigation and beyond. Nature neuroscience, 20(11), 1504.
Evans, G. (1982). The Varieties of Reference Oxford University Press Oxford.
Evans, J. S. B. T. (2010). Thinking Twice: Two Minds in One Brain. Oxford University Press.
Evans, J. St. B. T. (2003). In two minds: Dual-process accounts of reasoning. Trends in Cognitive Sciences, 7(10), 454–459. https://doi.org/10.1016/j.tics.2003.08.012
Evans, J., & Frankish, K. (2009). In Two Minds: Dual Processes and Beyond. Oxford University Press.
Falkenstein, M., Hohnsbein, J., Hoormann, J., & Blanke, L. (1991). Effects of crossmodal divided attention on late ERP components. II. Error processing in choice reaction tasks. Electroencephalography and clinical neurophysiology, 78(6), 447-455.
Faraji-Rad, A., & Pham, M. T. (2017). Uncertainty Increases the Reliance on Affect in Decisions. Journal of Consumer Research, 44(1), 1–21.
Feldman, H. and Friston, K. (2010). “Attention, Uncertainty, and Free-Energy”. Frontiers in human neuroscience 4, p. 215.
Feldman, H., & Friston, K. (2010). Attention, Uncertainty, and Free-Energy. Frontiers in Human Neuroscience, 4. https://doi.org/10.3389/fnhum.2010.00215
Ferguson, Don. "GIS for wilderness search and rescue." Proceedings of ESRI Federal User Conference. 2008.
Fernández Velasco, P. (2020). Disorientation and self-consciousness: A phenomenological inquiry. Phenomenology and the Cognitive Sciences. https://doi.org/10.1007/s11097-020-09659-1
Fernández Velasco, P., & Casati, R. (2020). Subjective disorientation as a metacognitive feeling. Spatial cognition & computation, 20(4), 281-305.
Fernandez Velasco, P., & Casati, R. (2020). The many faces of disorientation: a response to Daniel R. Montello. Spatial Cognition & Computation, 20(4), 314-327.
Fernández Velasco, P., & Casati, R. (2019). Disorientation and gis-informed wilderness search and rescue. In The Philosophy of GIS (pp. 241-251). Springer, Cham.
Fernández Velasco, P., & Casati, R. Making and Breaking Our Shared World: A Phenomenological Analysis of Disorientation as a Way of Understanding Collective Emotions in Distributed Cognition. The Politics of Emotional Shockwaves, 203.
Fernández Velasco, P., & Loev, S. (2020). Affective experience in the predictive mind: a review and new integrative account. Synthese, 1-36.
Fernández-Dols, J. M., & Ruiz-Belda, M.-A. (1997). Spontaneous facial behavior during intense emotional episodes: Artistic truth and optical truth. In The psychology of facial expression (pp. 255–274). Editions de la Maison des Sciences de l’Homme. https://doi.org/10.1017/CBO9780511659911.013
Filion, D. L., Dawson, M. E., Schell, A. M., & Hazlett, E. A. (n.d.). The Relationship Between Skin Conductance Orienting the Allocation of Processing Resources. Psychophysiology, 28(4), 410–424. https://doi.org/10.1111/j.1469-8986.1991.tb00725.x
FitzGerald, T. H., Friston, K. J., and Dolan, R. J. (2012). “Action-Specific Value Signals in Reward-Related Regions of the Human Brain”. Journal of Neuroscience 32.46, pp. 16417–16423.
Fleming, S. M., & Daw, N. D. (2017). Self-evaluation of decision-making: A general Bayesian framework for metacognitive computation. Psychological review, 124(1), 91.
Fleming, S. M., & Dolan, R. J. (2012). The neural basis of metacognitive ability. Philosophical Transactions of the Royal Society B: Biological Sciences, 367(1594), 1338–1349. https://doi.org/10.1098/rstb.2011.0417
Flournoy, T. (1893). Des phénomènes de synopsie (audition colorée): Photismes-schèmes visuels-personnifications. Paris: Félix Alcan.
Frankish, K., & Evans, J. S. B. T. (2009). In Two Minds: Dual Processes and Beyond. Oxford University Press.
Friedberg, M. S. di. (2017). Geographies of Disorientation. Routledge.
Frijda, N. H. (1986). The Emotions. Cambridge University Press.
Frijda, N. H. (2007). The Laws of Emotion. OCLC: 938467399. Lawrence Erlbaum Associates.
Frissen, I., Campos, J. L., Souman, J. L., & Ernst, M. O. (2011). Integration of vestibular and proprioceptive signals for spatial updating. Experimental brain research, 212(2), 163.
Friston, K. (2003). “Learning and Inference in the Brain”. Neural Networks 16.9, pp. 1325– 1352.
Friston, K. (2009). The free-energy principle: a rough guide to the brain?. Trends in cognitive sciences, 13(7), 293-301.
Friston, K., FitzGerald, T., Rigoli, F., Schwartenbeck, P., and Pezzulo, G. (2016). “Active Inference and Learning”. Neuroscience & Biobehavioral Reviews 68, pp. 862–879.
Friston, K., FitzGerald, T., Rigoli, F., Schwartenbeck, P., ODoherty, J., & Pezzulo, G. (2016a). Active inference and learning. Neuroscience & Biobehavioral Reviews, 68, 862–879. https://doi.org/10.1016/j.neubiorev.2016.06.022
Friston, K., FitzGerald, T., Rigoli, F., Schwartenbeck, P., ODoherty, J., & Pezzulo, G. (2016b). Active inference and learning. Neuroscience & Biobehavioral Reviews, 68, 862–879. https://doi.org/10.1016/j.neubiorev.2016.06.022
Friston, K., Mattout, J., and Kilner, J. (2011). “Action Understanding and Active Inference”. Biological cybernetics 104.1-2, pp. 137–160.
Froese, T. and Ikegami, T. (2013). “The Brain Is Not an Isolated “Black Box,” nor Is Its Goal to Become One”. Behavioral and Brain Sciences 36.3, pp. 213–214.
Fuchs, T. (2005). The phenomenology of body, space and time in depression. Comprendre, 15, 108-121.
Fuchs, T. (n.d.). The phenomenology of body, space and time in depression. 18.
Fulkerson, M. (2019). “Emotional Perception”. Australasian Journal of Philosophy, pp. 1– 15.
Furlanetto, T., Bertone, C., & Becchio, C. (2013). The bilocated mind: new perspectives on self-localization and self-identification. Frontiers in human neuroscience, 7, 71.
Gell, A. (1985). How to read a map: Remarks on the practical logic of navigation. Man, 271–286.
Gelman, Andrew, et al. Bayesian Data Analysis. Chapman and Hall/CRC, 2013. www.taylorfrancis.com, doi:10.1201/b16018.
Gerrans, P. (2015). “All the Self I Need”. Open MIND.
Gibson, J. J. (1979). The ecological approach to visual perception. Houghton, Mifflin and Company.
Gibson, James J. (1979). The Ecological Approach to Visual Perception. Houghton Mifflin.
Gibson, James Jerome. (1966). The Senses Considered as Perceptual Systems. Boston, USA: Houghton Mifflin.
Gill, C. (2010). Naturalistic Psychology in Galen and Stoicism. Oxford University Press.
Goldie, P. (2002). “Emotions, Feelings and Intentionality”. Phenomenology and the Cognitive Sciences 1.3, pp. 235–254.
Graziano, M.S., and Cooke, D.F. (2005). Parieto- frontal interactions, personal space, and defensive behavior. Neuropsychologia 44, 845–859.
Greifeneder, R., Bless, H., & Pham, M. T. (2011). When Do People Rely on Affective and Cognitive Feelings in Judgment? A Review. Personality and Social Psychology Review, 15(2), 107–141. https://doi.org/10.1177/1088868310367640
Griffiths, P. E., & Scarantino, A. (2005). Emotions in the Wild: The Situated Perspective on Emotion. In P. Robbins & M. Aydede (Eds.), The Cambridge Handbook of Situated Cognition. Cambridge University Press.
Gu, X., Hof, P. R., Friston, K. J., and Fan, J. (2013). “Anterior Insular Cortex and Emotional Awareness”. Journal of Comparative Neurology 521.15, pp. 3371–3388.
Gwizdka, J., & Spence, I. (2007). Implicit measures of lostness and success in web navigation. Interacting with Computers, 19(3), 357-369.
Hafting, Torkel, et al. “Microstructure of a Spatial Map in the Entorhinal Cortex.” Nature, vol. 436, no. 7052, Aug. 2005, pp. 801–06. www.nature.com, doi:10.1038/nature03721.
Halamish, V., McGillivray, S., & Castel, A. D. (2011). Monitoring one’s own forgetting in younger and older adults. Psychology and Aging, 26(3), 631–635. https://doi.org/10.1037/a0022852
Harsay, H. A., Spaan, M., Wijnen, J. G., & Ridderinkhof, K. R. (2012). Error Awareness and Salience Processing in the Oddball Task: Shared Neural Mechanisms. Frontiers in Human Neuroscience, 6.
Heft, H. (1996). The ecological approach to navigation: A Gibsonian perspective. In The construction of cognitive maps (pp. 105–132). Springer.
Heidegger, M (1962). Being and time, trans. Macquarrie, J, & Robinson, E. SCM Press, London.
Helm, B. W. (2009). “Emotions as Evaluative Feelings”. Emotion Review 1.3, pp. 248– 255.
Henderson, V. W., Mack, W., & Williams, B. W. (1989). Spatial disorientation in Alzheimer’s disease. Arch Neurol, 46(4), 391-4.
Heydrich, L., & Blanke, O. (2013). Distinct illusory own-body perceptions caused by damage to posterior insula and extrastriate cortex. Brain, 136(3), 790-803.
Hill, K. A. (1992). Spatial competence of elderly hunters. Environment and Behavior, 24(6), 779-794.
Hill, K. A. (1998). The psychology of lost. Lost Person Behavior. Ottawa: National SAR Secretariat, 1-16.
Hinde, R. A. (1985). Was ‘The expression of the emotions’ a misleading phrase? Animal Behaviour, 33(3), 985–992. https://doi.org/10.1016/S0003-3472(85)80032-4
Hinojosa, J. A., Carretié, L., Valcárcel, M. A., Méndez-Bértolo, C., & Pozo, M. A. (2009). Electrophysiological differences in the processing of affective information in words and pictures. Cognitive, Affective, & Behavioral Neuroscience, 9(2), 173-189
Hohwy, J. (2007). Functional integration and the mind. Synthese, 159(3), 315–328. https://doi.org/10.1007/s11229-007-9240-3
Hohwy, J. (2011). “Phenomenal Variability and Introspective Reliability”. Mind & Language 26.3, pp. 261–286.
Hohwy, J. (2012). Attention and conscious perception in the hypothesis testing brain. Frontiers in psychology, 3, 96.
Hohwy, J. (2012). Attention and Conscious Perception in the Hypothesis Testing Brain. Frontiers in Psychology, 3. https://doi.org/10.3389/fpsyg.2012.00096
Hohwy, J. (2013). The predictive mind. Oxford University Press.
Hohwy, J. (2017). Priors in perception: Top-down modulation, Bayesian perceptual learning rate, and prediction error minimization. Consciousness and Cognition, 47, 75-85
Hohwy, J., Roepstorff, A., and Friston, K. (2008). “Predictive Coding Explains Binocular Rivalry: An Epistemological Review”. Cognition 108.3, pp. 687–701.
Holyoak, K. J., & Morrison, R. G. (2013). The Oxford Handbook of Thinking and Reasoning. OUP USA.
Holyoak, K. J., & Morrison, R. G. (2013). The Oxford Handbook of Thinking and Reasoning. OUP USA.
Hommel, B. (2004). “Event Files: Feature Binding in and across Perception and Action”. Trends in Cognitive Sciences 8.11, pp. 494–500.
Horgan, T. and Tienson, J. (2002). “The Intentionality of Phenomenology and the Phenomenology of Intentionality”. Philosophy of Mind: Classical and Contemporary Readings, pp. 520–533.
Hornsby, Jennifer (2000) Personal and sub-personal: a defence of Dennett's early distinction. Philosophical Explorations: An International Journal for the Philosophy of Mind and Action 3 (1), pp. 6-24. ISSN 1386-9795.
Hughes, G., & Yeung, N. (2011). Dissociable correlates of response conflict and error awareness in error-related brain activity. Neuropsychologia, 49(3), 405-415.
Husserl, E. (1970). The crisis of European sciences and transcendental phenomenology: An introduction to phenomenological philosophy. Northwestern University Press.
Husserl, E. (1990). Ideas pertaining to a pure phenomenology and to a phenomenological philosophy: Second book studies in the phenomenology of constitution (Vol. 3). Springer Science & Business Media.
Husserl, E. (2012). Analyses Concerning Passive and Active Synthesis: Lectures on Transcendental Logic. Springer Science & Business Media.
Husserl, E. (2012). Ideas: General introduction to pure phenomenology. Routledge.
Hutchins, E. (1995). Cognition in the Wild. MIT Press.
Hutchins, E. (2005). Material anchors for conceptual blends. Journal of Pragmatics, 37(10), 1555–1577. https://doi.org/10.1016/j.pragma.2004.06.008
Hutchins, E. (2006). The distributed cognition perspective on human interaction. Roots of human sociality: Culture, cognition and interaction, 1, 375.
Hutchins, E. (2010a). Cognitive Ecology. Topics in Cognitive Science, 2(4), 705–715. https://doi.org/10.1111/j.1756-8765.2010.01089.x
Hutchins, E. (2010b). Enaction, imagination, and insight. In Enaction: Toward a New Paradigm for Cognitive Science. MIT Press.
Hutchins, E. (2014). The cultural ecosystem of human cognition. Philosophical Psychology, 27(1), 34–49. https://doi.org/10.1080/09515089.2013.830548
Hutchins, E., & Hazlehurst, B. (2002). Auto-Organization and emergence of shared language
Hutchins, E., & Johnson, C. M. (2009). Modeling the Emergence of Language as an Embodied Collective Cognitive Activity. Topics in Cognitive Science, 1(3), 523–546. https://doi.org/10.1111/j.1756-8765.2009.01033.x
Huth, J. E. (2013). The Lost Art of Finding Our Way. Harvard University Press.
Ingold, T. (2000). The perception of the environment: Essays on livelihood, dwelling and skill. Psychology Press.
Ionta, Silvio, et al. "Multisensory mechanisms in temporo-parietal cortex support self-location and first-person perspective." Neuron 70.2 (2011): 363-374.
Istomin, K. V., & Dwyer, M. J. (2009). Finding the way: A critical discussion of anthropological theories of human spatial orientation with reference to reindeer herders of northeastern Europe and western Siberia. Current Anthropology, 50(1), 29–49.
Jacoby, L. L., & Whitehouse, K. (1989). An illusion of memory: False recognition influenced by unconscious perception. Journal of Experimental Psychology: General, 118(2), 126.
James, W. (1884). “What Is an Emotion?” Mind 9.34, pp. 188–205.
James, W. (1890). The principles of psychology. New York : Holt. Retrieved from http://archive.org/details/theprinciplesofp01jameuoft
Joffily, M., & Coricelli, G. (2013). Emotional Valence and the Free-Energy Principle. PLOS Computational Biology, 9(6), e1003094. https://doi.org/10.1371/journal.pcbi.1003094
Johnston, R. S., & Dietlein, L. F. (1977). Biomedical results from Skylab. NASA SP-377. In Biomedical Results from Skylab (Vol. 377).
Julian, Joshua B., et al. “Place Recognition and Heading Retrieval Are Mediated by Dissociable Cognitive Systems in Mice.” Proceedings of the National Academy of Sciences, Apr. 2015, p. 201424194. www.pnas.org, doi:10.1073/pnas.1424194112.
Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux.
Kant, I. (2014). What does it mean to orient oneself in thinking? Daniel Fidel Ferrer, Verlag.
Kant, I., (1992 [1768]) “Concerning the Ultimate Ground of the Differentiation of Directions in Space”, in Walford, D. and Meerbote, R. (Trans. and Eds.), Theoretical Phylosophy, 1755–1770, The Cambridge Edition of the Works of Immanuel Kant (GuyerP. and WoodA.W., Eds.). Cambridge: Cambridge University Press, pp. 361–372
Keinath, Alex T., et al. “Environmental Geometry Aligns the Hippocampal Map during Spatial Reorientation.” Current Biology, vol. 27, no. 3, Feb. 2017, pp. 309–17. ScienceDirect, doi:10.1016/j.cub.2016.11.046.
Kenny, A. (1963). Action, Emotion and Will. London: Routledge & Kegan Paul.
Kensinger, E. A., & Corkin, S. (2004). Two routes to emotional memory: Distinct neural processes for valence and arousal. Proceedings of the National Academy of Sciences, 101(9), 3310–3315. https://doi.org/10.1073/pnas.0306408101
Kidd, C., Piantadosi, S. T., and Aslin, R. N. (2012). “The Goldilocks Effect: Human Infants Allocate Attention to Visual Sequences That Are Neither Too Simple nor Too Complex”. PloS one 7.5, e36399.
Kiverstein, J., Miller, M., and Rietveld, E. (2019). “The Feeling of Grip: Novelty, Error Dynamics, and the Predictive Brain”. Synthese 196.7, pp. 2847–2869.
Klatzky, R. L. (1998). Allocentric and Egocentric Spatial Representations: Definitions, Distinctions, and Interconnections. In C. Freksa, C. Habel, & K. F. Wender (Eds.), Spatial Cognition: An Interdisciplinary Approach to Representing and Processing Spatial Knowledge (pp. 1–17). Springer. https://doi.org/10.1007/3-540-69342-4_1
Klein, C. (2015). “What Pain Asymbolia Really Shows”. Mind 124.494, pp. 493–516.
Knierim, James J., and Derek A. Hamilton. “Framing Spatial Cognition: Neural Representations of Proximal and Distal Frames of Reference and Their Roles in Navigation.” Physiological Reviews, vol. 91, no. 4, Oct. 2011, pp. 1245–79. PubMed Central, doi:10.1152/physrev.00021.2010.
Koester, R. J. (2008). Lost Person Behavior: A search and rescue guide on where to look - for land, air and water (1st edition). Charlottesville, VA: dbS Productions LLC.
Koester, Robert J. Lost Person Behavior: A Search and Rescue. dbs Productions LLC, 2008.
Koopman, Bernard Osgood. Search and screening: general principles with historical applications. Vol. 7. New York: Pergamon Press, 1980.
Koriat, A. (2000). The feeling of knowing: Some metatheoretical implications for consciousness and control. Consciousness and cognition, 9(2), 149-171.
Koriat, A., & Nussinson, R. (2009). Attributing study effort to data-driven and goal-driven effects: Implications for metacognitive judgments. Journal of Experimental Psychology: Learning, Memory, and Cognition, 35(5), 1338–1343. https://doi.org/10.1037/a0016374
Kozuch, B. (2018). “No Pain, No Gain (in Darwinian Fitness): A Representational Account of Affective Experience”. Erkenntnis.
Kraut, R. E., & Johnston, R. E. (1979). Social and emotional messages of smiling: An ethological approach. Journal of Personality and Social Psychology, 37(9), 1539–1553. https://doi.org/10.1037/0022-3514.37.9.1539
Kriegel, U. (2014). “Towards a New Feeling Theory of Emotion”. European Journal of Philosophy 22.3, pp. 420–442.
Kubie JL, Levy ERJ, Fenton AA. 2019. Is hippocampal remapping the physiological basis for context? Hippocampus 543:23160. DOI: https://doi.org/10.1002/hipo.23160
Kubie, John L., and André A. Fenton. “Linear Look-Ahead in Conjunctive Cells: An Entorhinal Mechanism for Vector-Based Navigation.” Frontiers in Neural Circuits, vol. 6, 2012. Frontiers, doi:10.3389/fncir.2012.00020.
Lacewing, M. (2007). “Do Unconscious Emotions Involve Unconscious Feelings?” Philosophical Psychology 20.1, pp. 81–104.
Lang, P. J., Simons, R. F., Balaban, M., & Simons, R. (2013). Attention and Orienting: Sensory and Motivational Processes. Psychology Press.
Lange, C. G. (1887). Uber Gemütsbewegungen. Ihr Wesen Und Ihr Einfluß Auf Körperliche, Besonders Auf Krankhafte Lebenserscheinungen. Ein Medizinisch-Psychologische Studie. T. Thomas.
Lavrillier, A., & Gabyshev, S. (2017). An Arctic Indigenous Knowledge System of Landscape, Climate and Human Interactions. Evenki Reindeer Herders and Hunters.
Lavrillier, Alexandra. (2006). S’orienter avec les rivières chez les Évenks du Sud-Est sibérien. Un système d’orientation spatial, identitaire et rituel. Études Mongoles et Sibériennes, Centrasiatiques et Tibétaines, 36–37, 95–138.
Lazarus, R. S. (1991). Emotion and Adaptation. Oxford University Press.
Leary, M. R., Landel, J. L., & Patton, K. M. (1996). The motivated expression of embarassment following a self-presentational predicament. Journal of Personality, 64(3), 619–636. https://doi.org/10.1111/j.1467-6494.1996.tb00524.x
Levinson, S. C., & Levinson, S. C. (2003). Space in language and cognition: Explorations in cognitive diversity (Vol. 5). Cambridge University Press.
Levy, D. J. and Glimcher, P. W. (2012). “The Root of All Value: A Neural Common Currency for Choice”. Current Opinion in Neurobiology 22.6, pp. 1027–1038.
Li, X., Mou, W., & McNamara, T. P. (2012). Retrieving Enduring Spatial Representations After Disorientation. Cognition, 124(2), 143–155.
Lin, L., & Goodrich, M. A. (2010). A Bayesian approach to modeling lost person behaviors based on terrain features in wilderness search and rescue. Computational and Mathematical Organization Theory, 16(3), 300-323.
Lin, Lanny, and Michael A. Goodrich. “A Bayesian Approach to Modeling Lost Person Behaviors Based on Terrain Features in Wilderness Search and Rescue.” Computational and Mathematical Organization Theory, vol. 16, no. 3, Sept. 2010, pp. 300–23. Springer Link, doi:10.1007/s10588-010-9066-2.
Lin, Q., Zhang, D., Connelly, K., Ni, H., Yu, Z., & Zhou, X. (2015). Disorientation detection by mining GPS trajectories for cognitively-impaired elders. Pervasive and mobile computing, 19, 71-85.
Lindsay, D. S., & Kelley, C. M. (1996). Creating illusions of familiarity in a cued recall remember/know paradigm. Journal of Memory and Language, 35(2), 197-211.
Maier, M., Steinhauser, M., and Hubner, R. (2008). Is the error-related negativity amplitude related to error detectability? Evidence from effects of different error types. J. Cogn. Neurosci. 20, 2263–2273.
Mangan, B. (1993). Taking Phenomenology Seriously: The “Fringe” and Its Implications for Cognitive Research. Consciousness and Cognition, 2(2), 89–108. https://doi.org/10.1006/ccog.1993.1008
Mangan, B. (2000). What feeling is the “feeling of knowing?” Consciousness and Cognition: An International Journal, 9(4), 538–544. https://doi.org/10.1006/ccog.2000.0488
Mangan, B. (2001). Sensation’s ghost: The non-sensory “fringe” of consciousness. Psyche: An Interdisciplinary Journal of Research on Consciousness, 7(18), No Pagination Specified-No Pagination Specified.
Marr, D. (1982). Vision: A computational investigation into the human representation and processing of visual information. MIT Press. Cambridge, Massachusetts.
Martin, L. L., & Tesser, A. (2013). The Construction of Social Judgments. Psychology Press.
Mathews, A. and MacLeod, C. (2002). “Induced Processing Biases Have Causal Effects on Anxiety”. Cognition & Emotion 16.3, pp. 331–354.
McGlone, F., Wessberg, J., and Olausson, H. (2014). “Discriminative and Affective Touch: Sensing and Feeling”. Neuron 82.4, pp. 737–755.
Meinong, A. (1917). Uber emotionale Präsentation. A. Hölder.
Mendelovici, A. (2014). “Pure Intentionalism about Moods and Emotions”. Current Controversies in Philosophy of Mind, pp. 135–157.
Metzinger, T. (2017). The Problem of Mental Action. In T. Metzinger & W. Wiese (Eds.), Philosophy and Predicitive Processing.
Meylan, A. (2014). Epistemic Emotions: a Natural Kind? Philosophical Inquiries, 2(1), 173–190.
Michaelian, K., & Arango-Muñoz, S. (2014). Epistemic Feelings, Epistemic Emotions: Review and Introduction to the Focus Section. Philosophical Inquiries, 2(1), 97–122.
Michal, M., Reuchlein, B., Adler, J., Reiner, I., Beutel, M. E., Vögele, C., Schächinger, H., and Schulz, A. (2014). “Striking Discrepancy of Anomalous Body Experiences with Normal Interoceptive Accuracy in Depersonalization-Derealization Disorder”. PLOS ONE 9.2, e89823.
Millière, R. (2019). Are There Degrees of Self-Consciousness? Journal of Consciousness Studies, 26(3-4), 252-282.
Millikan, R. G. (1995). “Pushmi-Pullyu Representations”. Philosophical perspectives 9, pp. 185–200.
Monacelli, A. M., Cushman, L. A., Kavcic, V., & Duffy, C. J. (2003). Spatial disorientation in Alzheimer’s disease The remembrance of things passed. Neurology, 61(11), 1491-1497.
Moran, D., & Cohen, J. (2012). The Husserl Dictionary. Bloomsbury Publishing.
Mustonen, T., & Lehtinen, A. A. (2019). Lived displacement among the Evenki of Yiengra. International Journal of Critical Indigenous Studies, 12(2).
Nelson, R. K. (1969). Hunters of the Northern Ice. University of Chicago Press.
Nieuwenhuis, S., Ridderinkhof, K. R., Blom, J., Band, G. P., & Kok, A. (2001). Error-related brain potentials are differentially related to awareness of response errors: evidence from an antisaccade task. Psychophysiology, 38(5), 752-760.
Norman, E., Price, M. C., & Duff, S. C. (2010). Fringe Consciousness: A Useful Framework for Clarifying the Nature of Experience-Based Metacognitive Feelings. In Trends and Prospects in Metacognition Research (pp. 63–80). Springer, Boston, MA. https://doi.org/10.1007/978-1-4419-6546-2_4
O'Connell, R. G., Dockree, P. M., Bellgrove, M. A., Kelly, S. P., Hester, R., Garavan, H., ... & Foxe, J. J. (2007). The role of cingulate cortex in the detection of errors with and without awareness: a high‐density electrical mapping study. European Journal of Neuroscience, 25(8), 2571-2579
O’Keefe, J., and J. Dostrovsky. “The Hippocampus as a Spatial Map: Preliminary Evidence from Unit Activity in the Freely-Moving Rat.” Brain Research, vol. 34, 1971, pp. 171–75. APA PsycNET, doi:10.1016/0006-8993(71)90358-1.
Oman, C. M. (1987). The role of static visual orientation cues in the etiology of space motion sickness. In Proceedings of the symposium on vestibular organs and altered force environment (pp. 25-37).
Oman, C. M. (1990). Symptoms and signs of space motion sickness on Spacelab-1. Motion space sickness, 217-246.
Oman, C. (2007). Spatial Orientation and Navigation in Microgravity. In F. Mast & L. Jäncke (Eds.), Spatial Processing in Navigation, Imagery and Perception (pp. 209–247). Springer US. https://doi.org/10.1007/978-0-387-71978-8_13
Oman, C. M., Lichtenberg, B. K., Money, K. E., & McCoy, R. K. (1986). MIT/Canadian vestibular experiments on the Spacelab-1 mission: 4. Space motion sickness: symptoms, stimuli, and predictability. Experimental Brain Research, 64(2), 316-334.
Paillard J (1991) Motor and representational framing of space. In: Paillard J (ed) Brain and space. Oxford University Press, Oxford Pani JR, Dupree D (1994) Spatial reference system in the comprehension of rotational motion. Perception 23:929–946.
Parkinson, B. (1995). Ideas and Realities of Emotion. Routledge.
Parkinson, B., Fischer, A., & Manstead, A. S. R. (2005). Emotion in Social Relations: Cultural, Group, and Interpersonal Processes. Psychology Press.
Peacocke, C. (2019). The Primacy of Metaphysics. Oxford University Press.
Perrykkad, K. and Hohwy, J. (2020). “Fidgeting as Self-Evidencing: A Predictive Processing Account of Non-Goal-Directed Action”. New Ideas in Psychology 56, p. 100750.
Pessoa, L. (2005). “To What Extent Are Emotional Visual Stimuli Processed without Attention and Awareness?” Current opinion in neurobiology 15.2, pp. 188–196.
Pessoa, L. and Adolphs, R. (2010). “Emotion Processing and the Amygdala: From a ’low Road’ to ’Many Roads’ of Evaluating Biological Significance”. Nature Reviews Neuroscience 11, pp. 773–782.
Petkova, V. I., Björnsdotter, M., Gentile, G., Jonsson, T., Li, T. Q., & Ehrsson, H. H. (2011). From part-to whole-body ownership in the multisensory brain. Current Biology, 21(13), 1118-1122.
Pezzulo, G. and Cisek, P. (2016). “Navigating the Affordance Landscape: Feedback Control as a Process Model of Behavior and Cognition”. Trends in cognitive sciences 20.6, pp. 414–424.
Pezzulo, G., Rigoli, F., and Friston, K. (2015). “Active Inference, Homeostatic Regulation and Adaptive Behavioural Control”. Progress in neurobiology 134, pp. 17– 35.
Pezzulo, G., Rigoli, F., and Friston, K. J. (2018). “Hierarchical Active Inference: A Theory of Motivated Control”. Trends in cognitive sciences 22.4, pp. 294–306.
Ploner, M. (1999). “Pain Affect without Pain Sensation in a Patient with a Postcentral Lesion”, p. 4.
Power, C. C. P. M., Power, M., & Dalgleish, T. (2007). Cognition and Emotion: From Order to Disorder. Psychology Press.
Price, D. D. (2000). “Psychological and Neural Mechanisms of the Affective Dimension of Pain”. Science 288.5472, pp. 1769–1772.
Prinz, J. J. (2004). Gut Reactions: A Perceptual Theory of Emotion. Oxford University Press.
Prinz, J. (2005). “Are Emotions Feelings?” Journal of Consciousness Studies 12.8-9, pp. 9–25.
Prinz, J. J. (2006). Is emotion a form of perception?. Canadian Journal of Philosophy, 36(sup1), 137-160.
Prinz, J. (2010). For valence. Emotion Review, 2(1), 5-13
Proust, J. (2007). Metacognition and metarepresentation: is a self-directed theory of mind a precondition for metacognition? Synthese, 159(2), 271–295.
Proust, J. (2009). The Representational Basis of Brute Metacognition: A Proposal. In R. W. Lurz (Ed.), The Philosophy of Animal Minds (pp. 165–183). Cambridge University Press.
Proust, J. (2013). “Mental Acts as Natural Kinds”. Decomposing the Will, pp. 262–282.
Proust, J. (2013). The Philosophy of Metacognition: Mental Agency and Self-Awareness. Oxford University Press.
— 140 —
Proust, J. (2015). Feelings as Evaluative Indicators. In Open MIND. Frankfurt am Main: MIND Group. https://open-mind.net/epubs/feelings-as-evaluative-indicators2014a-reply-to-iuliia-pliushch/OEBPS/bk01.html
Proust, J. (2015). The representational structure of feelings. Open MIND. Frankfurt am Main: MIND Group.
Rajaram, S. (1993). Remembering and knowing: Two means of access to the personal past. Memory & cognition, 21(1), 89-102.
RANCK, J. B. “Head Direction Cells in the Deep Cell Layer of Dorsolateral Pre-Subiculum in Freely Moving Rats.” Electrical Activity of the Archicortex, 1985. CiNii, https://ci.nii.ac.jp/naid/10016937179/.
Ratcliffe, M. (2008). Feelings of Being: Phenomenology, Psychiatry and the Sense of Reality. Oxford University Press.
Rausch, M., Hellmann, S., & Zehetleitner, M. (2018). Confidence in masked orientation judgments is informed by both evidence and visibility. Attention, Perception, & Psychophysics, 80(1), 134-154.
Risko, E. F., & Gilbert, S. J. (2016). Cognitive Offloading. Trends in Cognitive Sciences, 20(9), 676–688. https://doi.org/10.1016/j.tics.2016.07.002
Roediger, H. L., & McDermott, K. B. (1995). Creating false memories: Remembering words not presented in lists. Journal of experimental psychology: Learning, Memory, and Cognition, 21(4), 803.
Rubins, J. L. and Friedman, E. D. (1948). “Asymbolia for Pain”. Archives of Neurology & Psychiatry 60.6, pp. 554–573.
Ruddle, R. A., Volkova, E., Mohler, B., & Bülthoff, H. H. (2011). The effect of landmark and body-based sensory information on route knowledge. Memory & Cognition, 39(4), 686–699.
Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39(6), 1161–1178. https://doi.org/10.1037/h0077714
Russell, J. A. (2003). Core affect and the psychological construction of emotion. Psychological Review, 110(1), 145–172. https://doi.org/10.1037/0033-295X.110.1.145
Safonova, T., & Sántha, I. (2013). Culture contact in Evenki land: A cybernetic anthropology of the Baikal region. Global Oriental.
Sartre, J. P. (1956). Being and nothingness, trans. Hazel E. Barnes. Philosophical Library, New York, 308.
Sava, E., Twardy, C., Koester, R., & Sonwalkar, M. (2016). Evaluating Lost Person Behavior Models. Transactions in GIS, 20(1), 38–53.
Scarantino, A. (2014). “The Motivational Theory of Emotions”. Moral Psychology and Human Agency, pp. 156–185.
Scarantino, A. (2017). Do Emotions Cause Actions, and If So How? Emotion Review, 9(4), 326–334. https://doi.org/10.1177/1754073916679005
Scheffers, M. K., & Coles, M. G. (2000). Performance monitoring in a confusing world: error-related brain activity, judgments of response accuracy, and types of errors. Journal of Experimental Psychology: Human Perception and Performance, 26(1), 141.
Schellenberg, S. (2007). Action and self-location in perception. Mind, 116(463), 603-632
Schupp, H. T., Markus, J., Weike, A. I., & Hamm, A. O. (2003). Emotional Facilitation of Sensory Processing in the Visual Cortex. Psychological Science, 14(1), 7–13. https://doi.org/10.1111/1467-9280.01411
Schwartz, B. L., & Metcalfe, J. (2011). Tip-of-the-tongue (TOT) states: retrieval, behavior, and experience. Memory & Cognition, 39(5), 737–749. https://doi.org/10.3758/s13421-010-0066-8
Schwenkler, J. (2014). Vision, self-location, and the phenomenology of the “point of view.” Noûs, 48(1), 137–155.
Serino, A., Alsmith, A., Costantini, M., Mandrigin, A., Tajadura-Jimenez, A., & Lopez, C. (2013). Bodily ownership and self-location: components of bodily self-consciousness. Consciousness and cognition, 22(4), 1239-1252
Seth, A. K. (2013). Interoceptive inference, emotion, and the embodied self. Trends in Cognitive Sciences, 17(11), 565–573. https://doi.org/10.1016/j.tics.2013.09.007
Seth, A. K. (2014). The cybernetic Bayesian brain. Open MIND. Frankfurt am Main: MIND Group.
Seth, A. K. and Friston, K. J. (2016). “Active Interoceptive Inference and the Emotional Brain”. Philosophical Transactions of the Royal Society B: Biological Sciences 371.1708, p. 20160007.
Seth, A. K. and Tsakiris, M. (2018). “Being a Beast Machine: The Somatic Basis of Selfhood”. Trends in cognitive sciences 22.11, pp. 969–981.
Seth, A. K., Suzuki, K., and Critchley, H. D. (2012). “An Interoceptive Predictive Coding Model of Conscious Presence”. Frontiers in Psychology 2.
Shalgi, S., Barkan, I., & Deouell, L. Y. (2009). On the positive side of error processing: error‐awareness positivity revisited. European Journal of Neuroscience, 29(7), 1522-1532.
Shemaykin, F. N. (1962). General problems of orientation in space and space representations. In Psychological science in the USSR (Vol. 1). US Office of Technical Reports Arlington.
Shirokogoroff, S. M. (1935). Psychomental complex of the Tungus (Vol. 1). Kegan Paul, Trench, Trubner London.
Sholl, M. Jeanne. “Cognitive Maps as Orienting Schemata.” Journal of Experimental Psychology: Learning, Memory, and Cognition, vol. 13, no. 4, 1987, pp. 615–28. APA PsycNET, doi:10.1037/0278-7393.13.4.615.
Siegel, A. W., & White, S. H. (1975). The development of spatial representations of large-scale environments. In Advances in child development and behavior (Vol. 10, pp. 9–55). Elsevier.
Solnit, R. (2005), A Field Guide to Getting Lost. London: Viking.
Sonnenfeld, J. (2002). Social dimensions of geographic disorientation in Arctic Alaska. Études/Inuit/Studies, 26(2), 157–173. JSTOR.
Sorce, J. F., Emde, R. N., Campos, J. J., & Klinnert, M. D. (1985). Maternal emotional signaling: Its effect on the visual cliff behavior of 1-year-olds. Developmental Psychology, 21(1), 195–200. https://doi.org/10.1037/0012-1649.21.1.195
Souman, J. L., Frissen, I., Sreenivasa, M. N., & Ernst, M. O. (2009). Walking straight into circles. Current biology, 19(18), 1538-1542.
Stanovich, K. E., & West, R. F. (2008). On the failure of cognitive ability to predict myside and one-sided thinking biases. Thinking & Reasoning, 14(2), 129–167.
Straus, E. W. (1966). Phenomenological psychology: the selected papers of Erwin W. Straus. Londres: Tavistock Publications.
structure. In A. Cangelosi & D. Parisi (Eds.), Simulating the evolution of language (pp. 279 305). London: Springer-Verlag.
Syrotuck, William G. Analysis of lost person behavior: An aid to search planning. Barkleigh Productions, 2000.
Tappolet, C. (2000). Emotions et Valeurs. FeniXX.
Taube, J. S., et al. “Head-Direction Cells Recorded from the Postsubiculum in Freely Moving Rats. I. Description and Quantitative Analysis.” Journal of Neuroscience, vol. 10, no. 2, Feb. 1990, pp. 420–35. www.jneurosci.org, doi:10.1523/JNEUROSCI.10-02-00420.1990.
Teroni, F. (2018). “Emotionally Charged—The Puzzle of Affective Valence”. Shadows of the Soul: Philosophical Perspectives on Negative Emotions, pp. 10–19.
Tolman, E. C. “The Determiners of Behavior at a Choice Point.” Psychological Review, vol. 45, no. 1, 1938, pp. 1–41. APA PsycNET, doi:10.1037/h0062733.
Tolman, Edward C. “Cognitive Maps in Rats and Men.” Psychological Review, vol. 55, no. 4, 1948, pp. 189–208. APA PsycNET, doi:10.1037/h0061626.
Tolstoy, Leo. Master and Man. Wildside Press, 2015.
Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. science, 185(4157), 1124-1131.
Tversky, B. (2000). Levels and structure of spatial knowledge. Cognitive mapping: Past, present and future, 24-43.
Tye, M. (1995). Ten Problems of Consciousness: A Representational Theory of the Phenomenal Mind. MIT Press.
Tye, M. (2008). The experience of emotion: An intentionalist theory. Revue internationale de philosophie, (1), 25-50
Uhelski, M. L. (2012). “Pain Affect in the Absence of Pain Sensation: Evidence of Asomaesthesia after Somatosensory Cortex Lesions in the Rat”, p. 8.
Van De Cruys, S. (2017). “Affective Value in the Predictive Mind”. Philosophy and Predictive Processing.
Van de Cruys, S. and Wagemans, J. (2011). “Putting Reward in Art: A Tentative Prediction Error Account of Visual Art”. i-Perception 2.9, pp. 1035–1062.
Van den Berg, K. H. (2018). Lostness in a Building Construction Project Simulation (Master's thesis).
Vasilevich, G. M. (1963). Drevnie geograficheskie predstavleniia èvenkov i risunok kart[Ancient Evenki geographical representations and mapping]. Изв. ВГО, 95(4), 306–319.
Wagner, A. (1999). Causality in Complex Systems. Biology and Philosophy, 14(1), 83–101. https://doi.org/10.1023/A:1006580900476
Waller, D., & Hodgson, E. (2006). Transient and Enduring Spatial Representations under Disorientation and Self-Rotation. Journal of Experimental Psychology: Learning, Memory, and Cognition, 32(4), 867–882.
Wang, R. F., & Spelke, E. S. (2000). Updating egocentric representations in human navigation. Cognition, 77(3), 215-250.
Wessel, J. R. (2012). Error awareness and the error-related negativity: evaluating the first decade of evidence. Frontiers in Human Neuroscience, 6, 88.
Wessel, J. R., Danielmeier, C., & Ullsperger, M. (2011). Error awareness revisited: accumulation of multimodal evidence from central and autonomic nervous systems. Journal of cognitive neuroscience, 23(10), 3021-3036.
Whittlesea, B. W. (1993). Illusions of familiarity. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19(6), 1235.
Whittlesea, B. W. A., & Williams, L. D. (1998). Why do strangers feel familiar, but friends don’t? A discrepancy-attribution account of feelings of familiarity. Acta Psychologica, 98(2), 141–165. https://doi.org/10.1016/S0001-6918(97)00040-1
Whittlesea, B. W. A., & Williams, L. D. (2001a). The discrepancy-attribution hypothesis: I. The heuristic basis of feelings and familiarity. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27(1), 3–13. https://doi.org/10.1037/0278-7393.27.1.3
Whittlesea, B. W. A., & Williams, L. D. (2001b). The discrepancy-attribution hypothesis: II. Expectation, uncertainty, surprise, and feelings of familiarity. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27(1), 14–33. https://doi.org/10.1037/0278-7393.27.1.14
Whittlesea, B. W., & Williams, L. D. (2000). The source of feelings of familiarity: the discrepancy-attribution hypothesis. Journal of Experimental Psychology: Learning, Memory, and Cognition, 26(3), 547.
Wiener, S. I., & Taube, J. S. (2005). Head Direction Cells and the Neural Mechanisms of Spatial Orientation. MIT Press.
Wiese, W. (2018). Experienced Wholeness: Integrating Insights from Gestalt Theory, Cognitive Neuroscience, and Predictive Processing. MIT Press.
Winkielman, P. and Berridge, K. C. (2004). “Unconscious Emotion”. Current directions in psychological science 13.3, pp. 120–123.
Winkielman, P., Berridge, K. C., and Wilbarger, J. L. (2005). “Unconscious Affective Reactions to Masked Happy versus Angry Faces Influence Consumption Behavior and Judgments of Value”. Personality and social psychology bulletin 31.1, pp. 121– 135.
Woollett, Katherine, and Eleanor A. Maguire. “Acquiring ‘the Knowledge’ of London’s Layout Drives Structural Brain Changes.” Current Biology, vol. 21, no. 24, Dec. 2011, pp. 2109–14. ScienceDirect, doi:10.1016/j.cub.2011.11.018.
Yeung, N., & Summerfield, C. (2012). Metacognition in human decision-making: Confidence and error monitoring. Philosophical Transactions: Biological Sciences, 367(1594), 1310–1321. JSTOR.
Yoder, Ryan M., et al. “Origins of Landmark Encoding in the Brain.” Trends in Neurosciences, vol. 34, no. 11, Nov. 2011, pp. 561–71. ScienceDirect, doi:10.1016/j.tins.2011.08.004.
I — INTRODUCTION Étudier la désorientation, c'est étudier comment, à travers notre corps, notre culture et la technologie, nous sommes connectés à notre environnement et ce qui se passe lorsque cette connexion est affaiblie ou rompue. Ce qui se passe, bien sûr, dépend à nouveau de notre environnement, de notre corps, de notre culture et de la technologie : ce n'est pas la même chose de se perdre dans les bois que dans un centre commercial, ce n'est pas la même chose d'être incapable d'interpréter une carte que de connaître un problème de GPS. Parfois, se perdre nous terrorise, nous paniquons et nous nous mettons à errer sans but. Dans d'autres cas, se perdre est un feu rouge qui nous dit de nous arrêter sur nos traces, car nous prenons le temps de reconsidérer nos repères et d'élaborer des stratégies d'orientation. Le monde qui nous entoure devient parfois étrange, inconnu ou dangereux lorsque nous sommes désorientés et, parfois, la désorientation est excitante et rafraîchissante - une invitation à explorer, à laisser derrière nous des désirs tenaces de contrôle et de certitude, et à adopter plutôt une relation plus spontanée avec notre environnement. Se perdre façonne notre conscience, non seulement en transformant notre perception du monde qui nous entoure, mais aussi en transformant notre sens de ce que nous sommes dans ce monde et des possibilités qui s'y offrent à nous. Il pourrait sembler à première vue que dans notre monde hautement technologique d'appareils GPS et d'applications de navigation, la désorientation est un phénomène rare qu'il faut considérer presque avec nostalgie. Loin de là, notre dépendance aveugle à la technologie peut elle-même être une cause de désorientation. Et notre monde moderne globalisé d'hyperinformation et de changements rapides peut être des plus désorientants. Les centres commerciaux et les grands centres de transport souterrains, malgré l'abondance des cartes et de la signalisation, sont parmi les endroits où il est le plus facile de se perdre, précisément à cause de leur absence de lieu ; un centre commercial ressemble à n'importe quel autre centre commercial et un point de vente d'une chaîne de café à n'importe quel autre point de vente de cette même chaîne de café. Dans ces environnements post-modernes, il n'y a pratiquement aucun sens du lieu qui puisse être développé et aucun point de repère distinct sur lequel fonder notre orientation. De nos jours, la désorientation est en fait omniprésente, et ce qui rend son étude difficile n'est pas sa rareté mais son omniprésence et sa nature multiforme. On peut étudier la désorientation à travers le prisme des neurosciences ou de la géographie humaine, et avec un objectif allant des opérations de recherche et de sauvetage à l'architecture des espaces publics. À ce jour, les innombrables domaines qui traitent de la désorientation n'ont pas été correctement reliés et nous ne disposons pas d'une image unifiée du phénomène, mais d'un ensemble de prises de vue fragmentées. Ce dont nous avons besoin, et c'est ce que je soutiendrai dans cette monographie, c'est d'une caractérisation unifiée à un niveau fondamental, à partir de laquelle une théorie de la désorientation peut émerger au profit des différentes disciplines concernées.
— 146 —
La désorientation est, au fond, un phénomène subjectif, et c'est en accordant l'attention nécessaire à ses aspects subjectifs que nous pouvons atteindre la caractérisation recherchée. C'est pourquoi une grande partie du présent travail est une enquête phénoménologique. Tout au long de la monographie, cette approche phénoménologique centrale est mise en communication avec d'autres disciplines, telles que les neurosciences, la géographie ou l'ethnographie. Cette approche interdisciplinaire s'inscrit dans la multiplicité de l'état actuel de la recherche sur la désorientation, et elle contraint et guide le travail conceptuel à différents niveaux explicatifs (par exemple phénoménal, neural, informatique, comportemental). En retour, le résultat est une théorie qui peut être adaptée pour fournir des explications et des prévisions pour un large éventail de disciplines. Une analyse phénoménologique des aspects négatifs et positifs de la désorientation révèle le lien entre notre sens de l'espace et notre sens du possible. Au cours de notre orientation quotidienne, il y a une intégration continue de représentations spatiales non-indexées de notre environnement (c'est-à-dire codées dans des cadres de référence allocentriques) et de représentations indexées de cet environnement (c'est-à-dire codées dans des cadres de référence égocentriques). Il en résulte que dans notre expérience standard d'un environnement bien connu, il y a toujours une configuration spatiale hors champ qui se profile au-delà de l'horizon de notre expérience et qui encadre cet horizon. Si nous connaissons bien Paris, lorsque nous nous promenons dans le Jardin de Luxembourg, nous aurons une idée de l'endroit où se trouvent les sorties du parc, et même de l'endroit où se trouve Notre Dame (et le Port Royal, Montparnasse, le Panthéon...), de sorte que nous pourrions la pointer sans imaginer explicitement une carte. Lorsque nous sommes désorientés, nous perdons ce sens de la localisation de notre environnement hors de vue par rapport à notre position actuelle. Le plus souvent, lors de la désorientation, nous nous retrouvons dans un espace à la fois appauvri et oppressant. Nous perdons non seulement le sens de la structure de notre environnement, mais aussi celui de nos possibilités au sein de cet environnement. Il en résulte souvent de la confusion, de l'anxiété et de l'impuissance. Cependant, il existe de rares occasions où la désorientation peut être une cause de joie et d'excitation. Dans ces cas-là, l'abandon de la structure standard avec laquelle nous appréhendons l'espace devient une invitation à établir des relations différentes avec notre environnement. Nous développons un sentiment de libération en nous laissant guider par les attraits du terrain. Cette double nature de la perte est à la fois une source d'émerveillement et un indice central dans notre compréhension du phénomène. Les analyses conceptuelles et phénoménologiques de cette monographie contribuent à faire avancer sa thèse centrale, à savoir que la meilleure façon de caractériser la désorientation est de la considérer comme un sentiment métacognitif. Ici, la façon de comprendre les sentiments métacognitifs (comme le sentiment de savoir, l'état du bout de la langue ou le sentiment de familiarité) est comme des expériences affectives concernant les propres états mentaux, processus ou capacités du sujet. Dans notre cas, un sous-système cognitif évalue et régule les processus de navigation active du sujet et, dans certaines conditions, cela se traduit par l'expérience de la désorientation. La théorie proposée est imbriquée dans deux cadres existants : Traitement prédictif et Cognition distribuée. Le traitement prédictif considère la prédiction comme l'activité fondamentale de l'esprit humain. Selon le traitement prédictif, le cerveau fait des hypothèses sur le monde en se basant sur l'expérience antérieure et ces hypothèses sont mises à jour avec les erreurs provenant des feuilles sensorielles dans une boucle de rétroaction qui fonctionne à différentes échelles de temps et à différents niveaux d'abstraction. L'idée centrale de la cognition distribuée est que les processus cognitifs sont répartis non seulement à l'intérieur de notre crâne, mais aussi dans un écosystème cognitif qui émerge de l'interaction de nombreux éléments tels que notre corps, un groupe d'agents ou des artefacts cognitifs. Une combinaison des deux nous donne une vue dans laquelle les agents opèrent au sein des écosystèmes cognitifs par le biais de processus cognitifs distribués qui
— 147 —
optimisent le succès des prédictions dans le temps. Les sentiments métacognitifs tels que la désorientation, que je défends, sont un élément central dans la régulation des différents éléments qui interviennent dans les processus cognitifs distribués, par le biais d'un contrôle évaluatif dans une boucle de mise à jour des prédictions. Il n'est pas nécessaire de donner plus de détails sur la théorie à ce stade. Il suffit de dire qu'avec cette théorie, nous voyons comment la désorientation ne fonctionne pas seulement comme une façon d'évaluer les représentations spatiales dans la tête d'un individu. La désorientation joue un rôle dans l'évaluation des processus de navigation qui peuvent être distribués au-delà de l'individu, sur des artefacts cognitifs (par exemple une carte ou un GPS), l'environnement (par exemple un système de signalisation de sentiers) et même des groupes de personnes (par exemple un guide ou un compagnon de chasse). Lorsqu'il y a des raisons de retirer la confiance dans le système de navigation en cours, une désorientation apparaît pour réguler la situation de manière dynamique. Parce que la désorientation se trouve à l'intersection de tant d'aspects cruciaux de la cognition humaine, lorsque nous jetons un coup d'œil sur le phénomène, nous en comprenons également certains aspects. Nous voyons également comment les sentiments (la désorientation en est un exemple) nous guident afin que nous puissions nous intégrer dans les vastes écosystèmes cognitifs que nous habitons. Nous voyons aussi comment les processus cognitifs distribués modulent notre expérience de l'espace et comment cette expérience de l'espace structure à son tour notre expérience du monde et de nous-mêmes. Et bien sûr, nous voyons ce qui se passe lorsque cette structure familière s'effondre.
II — MÉTHODOLOGIE Le présent travail s'engage dans de nombreux débats contemporains en philosophie analytique (par exemple, les liens entre l'auto-localisation et la conscience de soi), en anthropologie (par exemple, la théorie de la carte mentale par rapport à la théorie de la maîtrise pratique de l'orientation) et en sciences cognitives (par exemple, comment la cognition est distribuée au-delà de l'individu), pour ne citer que quelques disciplines. Les notions, les niveaux d'explication et même ce qui est considéré comme une explication valable diffèrent selon les disciplines. Par conséquent, passer d'une discipline à l'autre, c'est toujours marcher sur une fine ligne. Ici, la phénoménologie peut servir de moyen pour trouver un terrain d'entente. Une bonne compréhension du phénomène peut nous donner un aperçu et guider la recherche interdisciplinaire. Ceci, associé à une analyse conceptuelle rigoureuse, peut nous aider à faire progresser nos connaissances sur un sujet qui est destiné à nécessiter des efforts interdisciplinaires. L'analyse phénoménologique et conceptuelle de cette monographie s'appuie sur une série de rapports concernant des individus qui se perdent dans diverses situations. Certains de ces rapports sont historiques. Dans le chapitre 2, je m'appuie sur le corpus de rapports du XIXe siècle qu'Alfred Binet a rassemblés pour comprendre le phénomène de "retournement", une sous-classe d'expériences de désorientation dont la phénoménologie est très intéressante. Dans le chapitre 3, je m'appuie sur le rapport de l'anthropologue Maria Czaplicka, du début du XXe siècle, qui raconte comment son guide natif d'Evenki les a perdus dans la taïga sibérienne en pleine tempête de neige. D'autres rapports sont tirés de sources contemporaines. Kenneth Hill a recueilli une série de rapports de chasseurs perdus afin de catégoriser le comportement des sujets désorientés, ce qui éclaire une partie du travail conceptuel de cette monographie. Une autre série de rapports utiles provient d'astronautes qui se "retournent" (sur l'axe vertical au lieu de l'axe horizontal !) dans des environnements de microgravité, ce qui donne lieu à une phénoménologie des plus frappantes. La grande majorité des rapports contemporains proviennent d'une enquête que l'équipe de la désorientation de l'Institut Jean Nicod (dont je suis membre) a lancée en 2017 dans le but de
— 148 —
constituer un corpus d'expériences diverses de la désorientation. Dans ces rapports, les sujets exposent leurs expériences de désorientation, en donnant des détails sur la situation qui a provoqué l'épisode et les émotions qui y sont associées. Une dernière série de rapports est issue d'un travail ethnographique réalisé auprès des indigènes semi-nomades Evenki en Sibérie arctique et subarctique. En 2019, j'ai entrepris deux expéditions (une en hiver et une en été) dans la région d'Evenkia, dans le Kraï de Krasnoïarski, en Russie. Les chasseurs et les éleveurs de rennes d'Evenki sont réputés pour leur habileté à trouver leur chemin, et une grande partie de mon travail sur le terrain consistait à apprendre leurs méthodes de navigation et comment ces pratiques s'inscrivaient dans un écosystème cognitif plus large. Un autre aspect de la culture Evenki que je voulais apprendre était la phénoménologie de l'espace et du soi qu'ils expérimentent lors de la recherche de leur chemin. En particulier, j'étais intéressé par le sentiment qu'ils appellent "manakan", un sentiment d'autonomie et d'engagement avec leur environnement qui fait écho à la phénoménologie des manifestations positives de la désorientation. Les rapports que j'ai recueillis au cours de ces deux saisons de travail sur le terrain en Sibérie alimentent l'étude comparative du chapitre 3. Dans les premiers chapitres de la thèse, des rapports subjectifs sont souvent utilisés pour clarifier la phénoménologie de la désorientation et guider une analyse phénoménologique. Dans le chapitre dans lequel la revendication centrale est avancée, les rapports subjectifs sont également utilisés comme des scénarios qui délimitent ce qui devrait et ne devrait pas être considéré comme de la désorientation. L'utilisation des données de la première personne pour établir une théorie qui peut être appliquée pour expliquer les données de la troisième personne pose un défi, car il existe une lacune explicative lorsqu'il s'agit de relier les données de la première personne à celles de la troisième personne. Il y a au moins deux façons possibles de relever ce défi. La première est d'utiliser les données à la première personne provenant de la phénoménologie de la désorientation pour faire émerger des hypothèses vérifiables sur la cognition. La seconde est d'utiliser les données à la première personne provenant de la phénoménologie de la désorientation pour découvrir des liens entre des données empiriques existantes dispersées, puis d'utiliser ces données empiriques pour soutenir des hypothèses sur la cognition. L'approche que je suis dans ce travail est une combinaison des deux. En d'autres termes, l'analyse phénoménologique de la désorientation nous donne un aperçu du phénomène, ce qui nous aide à réinterpréter les travaux empiriques existants. Et quelles que soient les conclusions auxquelles ce processus aboutira, elles généreront des affirmations sur la cognition (par exemple, que la désorientation est un sentiment métacognitif) qui pourront être testées empiriquement à l'aide de données obtenues à la troisième personne.
III — APERÇU Pour arriver à la théorie unifiante souhaitée de la désorientation, je procède de la manière suivante : Le chapitre 1 présente les différentes disciplines qui traitent de la désorientation et plaide pour l'importance de développer une théorie unifiée du phénomène. J'utilise ensuite la recherche et le sauvetage comme étude de cas pour montrer comment cette théorie peut faire progresser la recherche dans différents domaines. Le chapitre 2 explore la phénoménologie de la désorientation et sa relation avec la conscience de soi. Il commence par examiner la littérature antérieure sur les liens entre l'auto-location et la conscience de soi et propose une distinction entre l'auto-location minimale (qui ne nécessite qu'un cadre de référence égocentrique) et l'auto-location intégrée (qui nécessite l'intégration de cadres de référence égocentriques et allocentriques). Le but de ce chapitre est d'utiliser cette distinction entre
— 149 —
l'auto-localisation minimale et l'auto-localisation intégrée pour approfondir notre compréhension de la désorientation spatiale. Une analyse phénoménologique des différents épisodes de désorientation révèle que dans les cas de désorientation paradigmatique, l'espace corporel se rétrécit et l'horizon de l'expérience devient plus incertain, ce qui conduit à l'anxiété et à un sentiment d'inconnu. La principale affirmation de ce chapitre est que pendant la désorientation, une déstabilisation de l'auto-localisation intégrée entraîne une diminution de la forme de conscience de soi. Le chapitre 3 explore les aspects positifs qui sont parfois associés aux expériences de désorientation. Pour tenter de donner un sens à ce paradoxe apparent, je me tourne vers un autre paradoxe apparent : le cas des éleveurs de rennes Evenki, dans le nord de la Sibérie, qui ne se seraient jamais perdus mais qui vivent le paysage d'une manière qui partage certains des aspects positifs des épisodes de désorientation. Ici, je m'appuie sur la littérature secondaire et sur mes propres interactions avec les éleveurs de rennes Evenki pour établir l'hypothèse suivante : une connaissance approfondie de leur paysage et une remarquable capacité d'orientation signifient que la navigation des Evenki n'est pas largement guidée par des représentations allocentriques. Cela signifie que l'intégration ou non de cadres de référence allocentriques et égocentriques n'est pas un facteur central dans les processus de navigation impliqués dans le comportement spatial typique des Evenki, et donc la déstabilisation de l'auto-localisation intégrée n'est pas quelque chose qui fait surface dans l'expérience consciente des éleveurs de rennes Evenki. Les Evenki ne se perdent jamais, non pas parce qu'ils savent toujours précisément où ils se trouvent par rapport à un grand cadre de référence allocentrique, mais parce que leurs interactions avec leur environnement sont structurées de telle manière qu'ils peuvent marcher en ne se fiant qu'à des cadres de référence égocentriques tout en étant capables de toujours trouver leur chemin. Cela signifie, à son tour, que leur expérience consciente de leur environnement (et d'eux-mêmes dans cet environnement) est largement déterminée par des cadres de référence égocentriques et non allocentriques, ce qui se traduit par une expérience de l'environnement à grande échelle comme étant sous-déterminée. Cela ne conduit cependant pas à l'anxiété, mais à un sentiment de nouveauté et de possibilité. En prenant cette analyse pour réfléchir aux aspects positifs des expériences de désorientation, je conclus que l'incertitude de l'espace environnant pendant la désorientation peut conduire non seulement à l'anxiété, mais aussi à une expérience de l'environnement comme nouvelle et pleine de possibilités. Le chapitre 4 s'appuie sur le corpus de rapports que l'équipe "Désorientation" a rassemblé pour avancer une caractérisation unifiée de la désorientation. Ce chapitre distingue d'abord la condition objective de la perte et la condition subjective de la désorientation. L'affirmation centrale du chapitre 4 (et de la monographie générale) est que la désorientation est un sentiment métacognitif qui évalue et régule les processus de navigation active. Cette caractérisation est compatible avec les données empiriques à la troisième personne et les données à la première personne sur l'expérience subjective de la désorientation, et elle sert de point central pour faire avancer une théorie du phénomène. Les chapitres 5 et 6 visent à offrir un compte rendu des expériences affectives dans le cadre du traitement prédictif. Le chapitre 5 commence par présenter un ensemble de caractéristiques communes aux expériences affectives (ou sentiments) qu'une théorie du traitement prédictif devrait viser à expliquer : les sentiments sont conscients, ils ont une valence, leur degré d'excitation varie, ils motivent un comportement et ce sont des états intentionnels avec des objets particuliers et formels. Je passe ensuite en revue les théories existantes sur les expériences affectives dans le cadre du traitement prédictif et je délimite deux familles de théories : Les théories d'inférence interceptive (qui affirment que les sentiments sont déterminés par des prédictions interceptives) et les théories de la dynamique des erreurs (qui affirment que les sentiments sont déterminés par les propriétés de la dynamique des erreurs). Le chapitre met ensuite en évidence les forces et les faiblesses de chaque famille de théories.
— 150 —
Le chapitre 6 s'appuie sur l'examen du chapitre précédent pour fournir une synthèse des familles existantes de théories de l'expérience affective dans le cadre du traitement prédictif : la théorie de l'inférence affective. La théorie de l'inférence affective affirme que la valence correspond au taux attendu de réduction des erreurs de prédiction. En retour, l'objet particulier d'un sentiment est l'objet dont on prédit qu'il sera la cause la plus probable des changements attendus du taux d'erreur de prédiction, et l'objet formel d'un sentiment est un modèle prédictif des changements attendus du taux d'erreur de prédiction causés par un objet particulier donné. Enfin, la théorie de l'inférence affective montre comment les expériences affectives biaisent la sélection des actions, orientant l'organisme vers l'homéostasie et vers des niveaux d'incertitude optimaux afin de minimiser l'erreur de prédiction dans le temps. Dans la dernière section du chapitre, j'utilise la théorie de l'inférence affective pour expliquer comment les sentiments métacognitifs émergent et comment ils guident le comportement. Le chapitre 7 prend la désorientation comme étude de cas pour élucider le rôle que les états affectifs jouent dans les processus cognitifs distribués. Le premier chapitre présente la cognition distribuée et les approches récentes de l'émotion. Ensuite, je présente quelques cas de cognition distribuée dans la navigation et j'examine le cas de la désorientation spatiale pour explorer le rôle de l'affect dans la navigation dans le contexte plus large de la cognition située. L'affirmation centrale de ce chapitre est que dans les processus cognitifs distribués, les états affectifs servent de forme de régulation évaluative qui contribue à la modulation d'un processus cognitif distribué et à l'éventuel passage entre les processus cognitifs distribués. Bien qu'ils soient liés par thème et qu'ils se réfèrent les uns aux autres, les chapitres 1 à 7 ne dépendent pas trop les uns des autres, en ce sens que les chapitres ultérieurs ne s'appuient pas fortement sur les arguments ou les revendications des chapitres précédents. Les chapitres 5 et 6, qui fonctionnent en tandem, font exception à cette règle. Ce niveau d'indépendance est particulièrement important car les chapitres 5-6 et 7 ne plaident pas directement en faveur du traitement prédictif ou de la reconnaissance distribuée, respectivement. Ils expliquent en quoi consistent les cadres théoriques et développent un compte rendu des états affectifs dans ces cadres. Je renvoie le lecteur à des sources qui défendent directement les cadres, mais si un lecteur a des raisons indépendantes contre l'un des cadres, il constatera qu'il peut lire le reste de la monographie sans rencontrer d'engagements forts en faveur de ce cadre. Le chapitre 8 rassemble les travaux développés dans le reste de la monographie. Le chapitre commence par synthétiser les récits des états affectifs développés dans les chapitres 5, 6 et 7 afin d'expliquer les sentiments métacognitifs. Ensuite, ce travail est utilisé pour rendre compte du fonctionnement de la désorientation, en ajoutant de la chair à l'os théorique introduit dans le chapitre 4. Cette nouvelle théorie de la désorientation est utilisée pour expliquer la phénoménologie de la désorientation, tant dans ses manifestations négatives que positives, telle qu'elle a été présentée dans les chapitres 2 et 3. Enfin, je montre comment la théorie peut contribuer à faire progresser les différentes disciplines concernées par le phénomène qui avait été introduit dans le chapitre 1. J'espère qu'après avoir lu le chapitre 8, le lecteur sera convaincu que cette monographie a accompli ce qu'elle s'était fixée au départ, à savoir fournir une théorie sur la manière dont nous, les humains, sommes connectés à notre environnement et sur ce qui se passe lorsque cette connexion s'effondre.
IV — UNE THÉORIE UNFIÉE DE LA DÉSORIENTATION L'affirmation centrale de cette monographie est que la désorientation est un sentiment métacognitif. L'objectif de ce chapitre est de s'appuyer sur cette affirmation pour développer un compte rendu de
— 151 —
la manière dont la désorientation apparaît et dont elle guide le comportement. Pour ce faire, j'établirai les liens nécessaires entre les autres chapitres de la monographie. Dans la section 5 de cet résumé, je ferai une synthèse du PP et de la cognition distribuée. Dans la section 6, cette synthèse sera utilisée pour développer la théorie de l'inférence affective (AIT) présentée au chapitre 6. La section 7 appliquera l'AIT au sentiment métacognitif de désorientation. La section 8 examinera comment la théorie unifiée émergente de la désorientation s'applique à la phénoménologie de la désorientation présentée au chapitre 2. La section 9 réexaminera les cas des aspects positifs de la désorientation et de la navigation Evenki (présentés au chapitre 3) à la lumière de la nouvelle théorie de la désorientation.
V — TRAITEMENT PRÉDICTIF ET COGNITION DISTRIBUÉE Comme nous l'avons vu au chapitre 5, selon le PP, l'organisme tentera de réduire l'erreur au fil du temps en actualisant les prévisions et en modifiant les schémas de gain de précision, ce qui implique des heuristiques (par exemple l'annulation de l'accélération optique), une dynamique corporelle passive (par exemple le mouvement bipède) et des pratiques culturelles - répartissant ainsi le travail entre le cerveau, le corps et le monde. Cette approche montre comment le traitement prédictif et la cognition distribuée peuvent fonctionner comme des cadres théoriques distincts mais fortement compatibles, ce qui est explicitement reconnu par les principaux partisans des deux cadres (voir Hutchins 2013 et Clark 2015 pour de bons exemples). Dans cette section, j'exposerai un défi important à la transition entre le PP et la cognition distribuée - le problème des pratiques culturelles entropiques - et je proposerai une solution à ce défi. La clé ici est l'idée que les pratiques culturelles tendent à diminuer l'entropie (et donc à augmenter la prévisibilité, car l'entropie est une mesure de l'imprévisibilité) à toutes les échelles d'un écosystème cognitif (Hutchins 2013). Le travail d'Edwin Hutchins fournit un inventaire complet (mais probablement pas exhaustif) des moyens par lesquels les pratiques culturelles y parviennent (Hutchins 2012), qui comprennent :
• réduction de la dimensionnalité - production d'une structure conceptuelle à partir d'assemblages complexes de matériaux éventuellement préconceptuels par la conjonction de caractéristiques (par exemple, la production d'une file d'attente à partir d'un groupe d'humains par la conjonction de la position dans la ligne avec l'ordre d'arrivée),
• filtrage - préservation de certaines caractéristiques ou éléments tout en en ignorant d'autres (par exemple, attirer notre attention sur les lignes blanches peintes au bord d'une route de montagne),
• la satisfaction des contraintes - l'accomplissement simultané de multiples restrictions qui
modifient la probabilité d'émergence de différentes configurations de pratiques culturelles (par exemple, lorsque l'on pratique le vélo, on observe un accomplissement simultané des contraintes du corps humain, de la mécanique du vélo et d'un riche code juridique et culturel, ce qui rend certaines manières de faire du vélo plus susceptibles d'émerger),
• la rétroaction positive modulée - le recyclage d'un sous-ensemble (filtré) de la sortie en tant
qu'entrée (par exemple, la réduction de la vitesse lorsque notre voiture s'approche dangereusement de la ligne blanche au bord de la route),
• superposition de structure - la projection d'une structure imaginaire sur des éléments d'un
monde perçu ou imaginé (par exemple, voir une constellation en regardant les étoiles).
— 152 —
• la cartographie à travers les espaces conceptuels - en combinant le filtrage avec la satisfaction des contraintes et la superposition afin de cartographier les modèles d'un espace conceptuel à l'autre (par exemple, traductions, comparaisons, analogies, métaphores...),
• et la conception - activités en dehors du flux de travail normal qui tentent de créer des
représentations explicites des pratiques de travail (par exemple, se mettre d'accord sur une chaîne de commandement au sein d'une organisation).
Hutchins explique comment ces méthodes de réduction de l'entropie sont présentes non seulement dans des pratiques culturelles relativement simples comme la file d'attente, mais aussi dans des pratiques très intellectuelles et complexes comme la recherche en physique quantique. Pour faire le lien entre ce qui précède et le PP, repensez au cas du joueur de base-ball (présenté au chapitre 5) qui utilise l'annulation de l'accélération optique (OAC) pour attraper la balle volante. Le corps du joueur doit être positionné de manière dynamique de telle sorte que la balle semble se déplacer à une vitesse constante dans le champ visuel. La façon dont cela se produit est d'attribuer un gain élevé à l'erreur associée à l'annulation de l'accélération verticale de la projection optique de la balle et à l'annulation d'autres erreurs (Clark 2015). Suivant l'inventaire de Hutchins, OAC est un exemple de rétroaction positive modulée, une pratique culturelle composite qui implique un filtrage en boucle dynamique. Le joueur de baseball filtre l'erreur sans rapport avec l'OAC et met à jour dynamiquement sa position afin de minimiser les changements de l'accélération perçue. En réfléchissant aux pratiques culturelles dans le cadre du PP, Kirchoff et Kiverstein soutiennent que "les contraintes qui découlent des pratiques culturelles influencent la manière dont la précision est pondérée dans un contexte donné et donc la manière dont l'incertitude est réduite au minimum" (p. 97, Kirchoff et Kiverstein, 2018). C'est exactement ce qui se passe dans les CAO. Les contraintes de la pratique culturelle influencent la façon dont la précision est pondérée (estimation du gain élevé pour l'erreur liée à la CAO) afin de réduire l'incertitude (sur la façon d'attraper la balle). Si nous suivons le principe de l'inférence active, lorsque nous disons que les pratiques culturelles tendent à augmenter la prévisibilité, nous ne parlons pas seulement de l'aspect descriptif des prédictions, mais aussi de leur aspect directif. En s'engageant dans la CAO, le joueur de base-ball ne rend pas seulement plus fiable la prédiction de l'endroit où la balle va tomber, mais aussi la prédiction que ledit joueur va attraper la balle. Les pratiques culturelles ont tendance à augmenter la prévisibilité (ou, par analogie, à diminuer l'incertitude ou l'entropie), mais ce n'est pas tout, car certaines pratiques culturelles ont plutôt tendance à diminuer la prévisibilité (ou, par analogie, à augmenter l'entropie). Appelons ce dernier type de pratiques culturelles "entropiques", par opposition aux pratiques culturelles "négentropiques" habituelles. C'est un point que Hutchins lui-même reconnaît, bien qu'il ne l'approfondisse pas. Hutchins ne fournit aucun inventaire des pratiques culturelles entropiques, il se contente de noter que celles-ci existent et que la cognition humaine traverse des cycles de désordre et de réorganisation à toutes les échelles de temps. Le seul exemple concret qu'il donne est que "dans la conduite d'une enquête scientifique, l'accumulation de désordres peut conduire à un réordonnancement conceptuel productif", et il note qu'"une secousse d'imprévisibilité est parfois nécessaire pour surmonter des structures conceptuelles stables mais inadéquates" (p. 321, Hutchins, 2012). Il est important de prendre le temps maintenant de discuter des pratiques culturelles négentropiques pour deux raisons. La première est qu'elles seront primordiales pour expliquer les aspects positifs de la désorientation dans la section 8.5. La seconde raison est que l'existence de pratiques culturelles qui diminuent réellement la prévisibilité pose un problème à la fois pour la cognition distribuée et pour le PP : pour la cognition distribuée, cela pose un problème parce que tout augmente ou diminue l'entropie, de sorte que le lien entre l'entropie et les pratiques culturelles devient beaucoup moins significatif une fois que nous disons simplement que certaines pratiques culturelles
— 153 —
diminuent l'entropie et d'autres non. Pour le PP, cela pose un problème parce que nous perdons le lien élégant entre l'augmentation de la prévisibilité et le principe PEM. Pour établir un lien complet entre la cognition distribuée et le PP, nous devons nous attaquer au problème des pratiques culturelles entropiques. Avant d'aller plus loin, il est important de préciser que le problème des pratiques culturelles entropiques n'est pas qu'une pratique donnée n'augmente pas la prévisibilité dans certains cas. Les écosystèmes cognitifs augmentent simplement la probabilité d'apparition de certains modèles, c'est-à-dire que l'interaction entre l'écosystème cognitif et le modèle d'activité qui en résulte est non linéaire. Dans ces systèmes dynamiques, il n'est pas logique de se fixer sur une causalité régulière (c'est-à-dire une causalité dans le sens où si A cause B, alors A doit toujours être suivi par B). Une notion de causalité régulière ne peut être définie de manière significative pour les systèmes sans interactions linéaires entre leurs variables (pour une discussion mathématique, voir Wagner 1999). Lorsque nous discutons de pratiques culturelles, nous devrions suivre une notion probabiliste de causalité. Utiliser cette notion, pour dire que les pratiques culturelles négentropiques augmentent la prévisibilité, revient à dire que les pratiques culturelles négentropiques rendent une augmentation de la prévisibilité plus probable. Le problème des pratiques entropiques ne se pose que lorsque nous trouvons certaines pratiques culturelles qui ont tendance à diminuer la prévisibilité au lieu de l'augmenter. Pour résoudre le problème des pratiques culturelles entropiques, nous devons attirer l'attention sur la façon dont la cognition est largement répartie dans l'espace et dans le temps. Un premier pas dans cette direction consiste à reconnaître comment les pratiques culturelles permettent l'apprentissage. Les jeux en sont un excellent exemple : ils nous fournissent une entropie initiale élevée (par exemple, nous ne répondons pas correctement aux questions d'un jeu-questionnaire), nous progressons lentement dans la réduction de cette entropie (par exemple, nous répondons aux questions avec de plus en plus de précision) et nous apprenons dans le processus (par exemple, nous améliorons nos connaissances générales). L'apprentissage est un point sur lequel Clark insiste dans son effort de rapprochement entre les pratiques culturelles et le PP : "en concevant et en redessinant sans cesse nos propres environnements, en les peuplant de nouveaux livres, peintures, théories, jeux et pratiques, nous, les humains, déplaçons continuellement les poteaux d'objectif pour notre propre apprentissage basé sur la prédiction" (p. 531, Clark 2018). L'idée est qu'en étant ancrés dans une culture, nous apprenons à valoriser (c'est-à-dire à assigner une grande précision) certaines pratiques culturelles qui nous aident (en permettant l'apprentissage) à atteindre notre objectif de réduction des erreurs au fil du temps. Néanmoins, les pratiques qui permettent d'apprendre ne semblent pas tout à fait correspondre à ce que Hutchins appelle "une secousse d'imprévisibilité" et "un désordre qui s'accumule". Hutchins pensait notamment aux pratiques scientifiques visant à saper les théories existantes. Un autre bon exemple de ces pratiques culturelles entropiques perturbatrices en dehors du contexte scientifique est la technique de détournement de l'Internationale lettrée, qui consiste à détourner les éléments esthétiques de la culture dominante et à les transformer de manière subversive (par exemple en réorganisant les lettres dans un slogan ou un nom de marque) (Debord et Wolman 1956). La plupart des pratiques de protestation illustrent également des pratiques culturelles entropiques perturbatrices. Le problème avec ces pratiques culturelles perturbatrices est qu'elles ne peuvent pas être facilement expliquées en faisant appel à l'apprentissage. Il n'y a pas de moyen clair pour qu'elles permettent d'améliorer progressivement le succès des prédictions. Pour couvrir à la fois les types d'apprentissage et de perturbation des pratiques entropiques, je vais avancer et défendre la conjecture suivante : les pratiques culturelles entropiques pourraient diminuer la prévisibilité à court terme et l'augmenter à long terme. C'est certainement ce qui se passe dans le cas de l'apprentissage et, je soutiendrai, c'est également ce qui se passe avec les pratiques culturelles
— 154 —
perturbatrices, comme le cas de la recherche scientifique mentionné par Hutchins. Une série de pratiques culturelles entropiques déstabilisent une structure théorique (par exemple en créant des problèmes que la structure ne peut pas assimiler), ce qui entraîne un réarrangement conceptuel. Les pratiques perturbatrices sont utiles lorsque les écosystèmes cognitifs tournent autour de pratiques culturelles négentropiques qui augmentent la prévisibilité de manière globalement sous-optimale. En d'autres termes, une pratique culturelle négentropique donnée survit parce qu'elle parvient à accroître la prévisibilité, mais il existe une possibilité latente de nouvelles pratiques culturelles négentropiques qui permettront d'améliorer la prévisibilité. Ensuite, un rôle probable des pratiques perturbatrices serait de déstabiliser l'écosystème cognitif jusqu'à ce qu'une nouvelle pratique culturelle négentropique émerge et fasse un meilleur travail pour diminuer la prévisibilité. Le rôle des pratiques culturelles perturbatrices est précisément d'induire ces périodes de désordre fructueux32. Porter une attention particulière à la dynamique des écosystèmes cognitifs et à la manière dont les pratiques culturelles opèrent à différentes échelles, tant dans l'espace que dans le temps, nous permet d'aborder le problème des pratiques entropiques. Toutes les pratiques culturelles tendent à réduire l'entropie. La mise en garde est que différentes pratiques réduisent l'entropie à différentes échelles de temps. Les pratiques culturelles négentropiques (comme celles qui figurent dans l'inventaire détaillé de Hutchins) le font à des échelles temporelles relativement courtes. Les pratiques culturelles entropiques (pratiques perturbatrices et pratiques d'apprentissage) le font à des échelles de temps plus longues. Ensemble, elles contribuent à accroître la prévisibilité dans le temps dans un écosystème cognitif donné. Cette augmentation de la prévisibilité dans le temps contribue aux activités de minimisation des erreurs des humains opérant au sein dudit écosystème cognitif. Cette caractérisation des pratiques culturelles fournit le lien recherché entre la PP et la cognition distribuée.
VI — THÉORIE DE L’INFERENCE AFFECTIVE Au chapitre 6, l'ACI a expliqué comment les sentiments métacognitifs émergent de l'évaluation et de la régulation d'un processus cognitif et comment ils guident le comportement :
• La valence phénoménale correspond au taux attendu associé au processus cognitif. • L'excitation ressentie correspond au changement de rythme attendu associé au processus
cognitif. • L'objet particulier correspond au processus cognitif dont on déduit qu'il provoque des
changements dans la dynamique des erreurs.
32 Ici, la question se pose de savoir pourquoi il y aurait des pratiques négentropiques sous-optimales en premier lieu. Il y a
deux raisons pour lesquelles des pratiques négentropiques sous-optimales pourraient apparaître, et toutes deux sont liées au contexte dans lequel elles se développent. La première raison est la dépendance au cheminement, une notion populaire en sociologie historique (Mahoney 2000) : les pratiques culturelles se développent de manière contingente, à partir des vicissitudes de leur époque et de leur environnement. Cela peut conduire à des pratiques qui sont juste "assez bonnes" au niveau local (pour une défense de ce "principe assez bon" dans les pratiques linguistiques, voir Ferreira et Patson 2007). La deuxième raison connexe est que des pratiques négentropiques peuvent apparaître dans une situation particulière où elles réduisent efficacement l'entropie et persistent ensuite même lorsque leur efficacité a considérablement diminué au fil du temps (par exemple, pensez à la façon dont les utilisateurs pourraient s'en tenir à des programmes informatiques qu'ils connaissent bien même si de meilleurs programmes ont déjà été développés). Ces pratiques sous-optimales persistent parce que les pratiques culturelles sont à la fois auto-renforcées (par exemple, plus les gens respectent les files d'attente, plus ils sont susceptibles d'en former) et se renforcent mutuellement (par exemple, "les façons de parler de 'premier de la file', 'suivant', 'arrière de la file', etc. sont des pratiques discursives qui entrent dans des relations de renforcement mutuel avec la conception de la matrice spatiale linéaire comme une file d'attente", p.8, Hutchins, 2013), et, ce qui est le plus important pour notre discussion actuelle, parce qu'il pourrait être impossible de passer à une pratique négentropique plus efficace sans d'abord subir une période de diminution de l'efficacité en termes de prévisibilité.
— 155 —
• La performance du processus cognitif est modélisée par un objet formel, qui est à la fois un modèle descriptif et une politique d'action directive.
• Par son objet formel, le sentiment métacognitif guide l'action (tant mentale que physique) en transformant le paysage de l'affordance (cognitive) du sujet et en biaisant la concurrence entre les actions potentielles.
Au chapitre 7, j'ai présenté un compte rendu parallèle sur la manière dont les états affectifs évaluent et régulent les processus cognitifs distribués de manière diachronique et synchrone. Diachroniquement, les états affectifs peuvent conduire à la promotion de certains processus cognitifs (par exemple, la honte favorisant l'apprentissage des compétences de navigation chez les natifs Inupiat). De manière synchrone, l'évaluation de différents modes de cognition répartis à différentes échelles (M1-M4) peut conduire à
• la modulation (d'un seul processus cognitif, et donc dans un seul mode), • ou au désengagement (c'est-à-dire au passage entre les processus cognitifs, et le plus souvent
entre les modes de traitement M1, M2, M3 et M4).
La discussion dans la section précédente permettra de voir comment les approches des chapitres 6 et 7 se rejoignent. Examinons d'abord le rôle diachronique des états affectifs dans l'écosystème cognitif. Nous avons vu que les écosystèmes cognitifs fournissent des contraintes qui rendent certains assemblages cognitifs plus susceptibles d'émerger. Ici, les pratiques émotionnelles de fond (par exemple, la honte) fonctionnent comme des contraintes qui rendent certaines pratiques culturelles plus susceptibles d'émerger (par exemple, l'apprentissage des compétences de navigation). En termes de PP, elles favorisent certains schémas de pondération de précision. Dans le cas de l'Inupiat, l'attente d'une éventuelle honte fait en sorte que, toutes choses égales par ailleurs, l'engagement dans des pratiques d'apprentissage (entropiques) et de navigation (négentropiques) devient le moyen le plus efficace de réduire les erreurs au fil du temps. Inversement, l'apprentissage des pratiques culturelles permettra aux agents d'affiner leur mécanisme affectif en apprenant quels sont les meilleurs indices pour prédire la dynamique des erreurs et comment modéliser cette dynamique de manière à l'adapter à l'écosystème cognitif particulier (par exemple, ce dont il faut avoir peur, comment exprimer sa tristesse, que faire lorsque l'on se méfie de quelqu'un...). Nous n'avons pas besoin de l'AIT pour comprendre comment les états affectifs contraignent les processus cognitifs de manière diachronique, car l'AIT vise principalement les liens synchroniques entre les états affectifs et les processus cognitifs. L'AIT permet de comprendre comment les états affectifs évaluent de manière synchrone les processus cognitifs répartis à différentes échelles. Lorsqu'un processus cognitif est en cours, quelle que soit l'échelle de distribution, il est associé à certaines dynamiques d'erreurs attendues. Par conséquent, un sentiment métacognitif apparaîtra lorsqu'il y aura des changements substantiels dans la dynamique attendue, et l'objet particulier correspondra au processus cognitif en question. Imaginez que vous observez les étoiles et que vous ne parvenez pas à nommer ce qui ressemble à une constellation familière - vous obtiendrez probablement un état du bout de la langue (ToT) orienté vers vos connaissances astronomiques. Si vous essayez ensuite d'utiliser vos doigts pour mesurer la distance angulaire entre les étoiles mais que vous vous souvenez mal de la méthode et que les choses ne semblent pas s'additionner, vous pourriez avoir un sentiment d'erreur dirigé vers cette compétence incarnée. Si vous sortez votre téléphone et essayez d'utiliser une application d'observation des étoiles, et que l'application ne fonctionne pas, votre sentiment métacognitif (probablement l'inverse de la confiance) sera cette fois-ci dirigé vers l'application. Le croquis ci-dessus illustre bien la façon dont les sentiments métacognitifs modulent les processus cognitifs qui se répartissent dans le cerveau, le corps et le monde, et comment ils suscitent le désengagement et l'engagement dans les processus cognitifs à diverses échelles. En termes d'AIT, un processus cognitif distribué (c'est-à-dire l'objet particulier) sera associé à une certaine dynamique
— 156 —
d'erreur attendue, qui donnera naissance à un objet formel qui modélisera la dynamique (par exemple, en les classant comme un état de FdF, un sentiment d'erreur, etc) et transformera le paysage de l'affordance du sujet (via des politiques d'action). ) et transforme le paysage financier du sujet (par le biais de politiques d'action). Cette transformation du paysage financier entraînera une modulation (par exemple, guider la recherche mentale du nom de la constellation, alterner la façon dont nous encadrons les étoiles avec nos doigts, modifier les réglages du téléphone) et, en cas d'échec, un désengagement éventuel (par exemple, abandonner la recherche mentale et essayer de mesurer les étoiles avec nos doigts, ou abandonner cette méthode et retirer le téléphone). Comme nous l'avons vu au chapitre 6, les objets formels comprennent à la fois des politiques d'action et des modèles de processus cognitifs, de sorte qu'un changement de politiques d'action s'accompagnera souvent d'un changement dans la façon dont les dynamiques attendues sont modélisées. Imaginons qu'un sujet soit demandé pour la capitale de la Serbie. Au début, elle a l'impression de connaître la réponse (sentiment de savoir) et essaie de la rappeler. Elle n'y parvient pas et subit un état de ToT. Après de nombreux efforts mentaux, elle finit par abandonner par frustration. Elle connaîtra probablement non seulement un changement dans les politiques d'action (récupération et désengagement ultérieur de la récupération), mais aussi un changement dans le sentiment métacognitif vécu (de la sensation de savoir à l'état de FdF et de cet état de FdF à la sensation de ne pas savoir). Si un objet formel fournit un modèle de la dynamique (connaître la réponse) et une politique d'action (récupération) pour réguler cette dynamique, l'échec de cette politique d'action à réguler la dynamique sera considéré comme le soutien d'une hypothèse différente (que le sujet ne connaît pas la réponse), conduisant à un nouvel objet formel à la fois sous la forme d'une nouvelle politique d'action (arrêt du processus de récupération) et d'un nouveau modèle descriptif de la dynamique (le sujet ne connaissant pas la réponse). Le scénario ci-dessus montre également comment les sentiments métacognitifs émergent successivement afin de guider les processus cognitifs distribués de manière dynamique et flexible. Un aspect intéressant de cette dynamique est la fréquence à laquelle un sujet passe d'un processus cognitif à un autre, plutôt que de se désengager progressivement d'un processus pour s'engager dans un autre. Cet aspect est particulièrement important pour nos objectifs car, dans des conditions expérimentales, il a été signalé que des sujets désorientés passaient de systèmes de représentation spatiale en ligne à des systèmes hors ligne (Waller et Hodgson 2006). En PP, les commutations sont traditionnellement conçues comme des changements rapides d'hypothèses, l'étude de cas standard étant les commutations perceptuelles pendant la rivalité binoculaire (Hohwy et al. 2008). Bien entendu, les hypothèses comprennent non seulement des prédictions de stimuli perceptifs, mais aussi des prédictions proprioceptives (c'est-à-dire des actions). Un passage entre des processus cognitifs doit alors être conçu comme un passage entre des politiques d'action entraînant un réarrangement rapide du paysage de l'accessibilité (et éventuellement un changement des aspects descriptifs de l'objet formel également). Pour clarifier les choses, il n'est pas nécessaire de spéculer sur une correspondance étroite et bidirectionnelle entre les processus cognitifs distribués et les politiques d'action. Pour les besoins de la présente étude, il suffit de dire que lorsqu'un sentiment métacognitif suscite un changement entre des processus cognitifs, il le fait par le biais d'un changement de politiques d'action. Une question intéressante est de savoir quels sentiments, et dans quelles conditions, auront tendance à provoquer un changement plutôt qu'un désengagement progressif. Il s'agit essentiellement d'une question empirique, mais nous pouvons avancer la conjecture théorique suivante : Les commutations se produiront lorsque l'organisme devra s'adapter à ce qui devrait être une dynamique d'erreur en évolution rapide. En termes d'ACI, la dynamique d'erreur à évolution rapide attendue correspond à un changement de taux attendu élevé, c'est-à-dire à des niveaux d'éveil élevés. Par
— 157 —
conséquent, plus son niveau d'éveil est élevé, plus un état affectif à valence négative33 sera susceptible de provoquer des changements entre les politiques d'action avec le changement correspondant entre les processus cognitifs distribués. Cette idée est indirectement soutenue par la théorie de la concurrence biaisée par l'excitation (Mather et Sutherland 2011), qui propose que lorsqu'un signal est considéré comme hautement prioritaire, l'excitation amplifie ce signal, ce qui entraîne un avantage concurrentiel dans la concurrence biaisée (voir Lee et al 2014 pour une étude neuroscientifique à l'appui de la théorie de la concurrence biaisée par l'excitation). Lorsque nous plaçons la théorie de la concurrence biaisée par l'excitation en termes de PP et de concurrence par les moyens, l'excitation accélère le processus de sélection entre les hypothèses concurrentes, ce qui entraîne des commutations plutôt que des transformations progressives (pour une étude montrant la corrélation entre l'excitation et le taux de commutation dans les expériences de rivalité binoculaire, voir Carter et al 2007). Une conclusion connexe est que l'excitation augmente le taux d'apprentissage (Nassar et al 2012), ce qui renforce la compréhension de l'excitation par l'AIT, car on s'attendrait à ce que le cerveau mette à jour l'hypothèse à la lumière du revenu sensoriel à un rythme plus élevé plus la dynamique des erreurs devrait changer rapidement. A un niveau fondamental, les états affectifs contribuent à l'harmonisation au sein d'une pratique culturelle. Pour revenir au cas du CAO, un sentiment subtil d'erreur fonctionnera comme un mécanisme de rétroaction pour s'assurer que le joueur minimise les changements d'accélération optique et attrape la balle. L'objectif d'attraper la balle se traduira par une prédiction de stabilité de l'accélération optique, et un sentiment apparaîtra si des erreurs sont attendues en rapport avec cette prédiction, régulant ainsi le mouvement pour s'assurer que la dynamique du joueur est couplée à celle de la balle. Ce que le cas OAC révèle, c'est qu'il n'est pas nécessaire que les erreurs grimpent aux niveaux supérieurs de la hiérarchie. Une fois que les bonnes prédictions ont été établies à un niveau élevé de la hiérarchie de prédiction (par exemple, l'objectif de minimiser l'accélération optique), un sentiment d'erreur va surveiller la situation de manière dynamique et guider l'organisme dans les ajustements nécessaires en modifiant subtilement le schéma des pondérations de précision. Si l'on s'attend à des changements rapides et négatifs du taux, davantage de ressources seront employées pour contrôler la dynamique des erreurs, et un changement de la politique d'action deviendra plus probable. Là encore, l'éveil influencera le comportement en fonction du contexte. Dans une journée venteuse, une forte excitation entraînera l'utilisation de plus de ressources, tant que la balle est perçue comme pouvant être attrapée. Et si la balle va trop vite, cela entraînera très vite l'abandon de la stratégie de la CAO, avec la disparition correspondante de l'accessibilité de la balle à l'attrapage, comme le montrent de récents travaux expérimentaux (Postma et al 2018). Un dernier élément développé dans le chapitre 7 que nous devons présenter en termes d'ACI concerne les théories du double processus, car l'affirmation était que les états affectifs provoquent des commutations non seulement entre des processus cognitifs répartis à différentes échelles, mais aussi entre des modes de cognition "rapides" et "lents" (M1>M2, M4>M3, etc.). Dans le PP, les modes de cognition rapides et lents sont conceptualisés au sein de la hiérarchie prédictive. Les objectifs de haut niveau contextualisent les réponses moins flexibles aux niveaux inférieurs (Pezzulo et al 2015). La sélection de modes de cognition lents ou rapides dépend de la précision des pondérations au sein de la hiérarchie. Plus le gain de précision attribué aux niveaux supérieurs (inférieurs) est important, plus il est probable que des modes de cognition lents (rapides) soient sélectionnés. "Le résultat est un continuum entre l'action orientée vers un but et l'action habituelle qui dépend du poids relatif attribué aux différentes couches. Des habitudes peuvent également apparaître dans ce schéma, lorsque les couches inférieures acquièrent une précision suffisante pour devenir essentiellement imperméables aux influences des couches supérieures" (p.14, Pezzulo 2017). Le lien avec l'ACI est assez simple : les états affectifs entraînent une réorganisation rapide de la pondération de la précision (c'est-à-dire une transformation du paysage de l'accessibilité). Cette 33 Si l'on prévoit une augmentation rapide du taux (c'est-à-dire une forte excitation et une valence positive), il faut alors
poursuivre le processus en cours.
— 158 —
conception des commutations intermodes est conforme à l'idée du PP selon laquelle les modèles de pondération de précision recrutés de manière contextuelle "accomplissent une forme de sélection d'ensemble ou de changement de stratégie" (p. 14, Clark 2015). Là encore, l'excitation joue un rôle important. Plus le taux de changement attendu est élevé, plus la réponse sélectionnée aura tendance à être basse le long de la hiérarchie, car les niveaux inférieurs opèrent dans des délais plus courts. Ceci est cohérent avec une grande partie du travail expérimental sur l'éveil. Par exemple, Wichary et ses collègues ont constaté que l'excitation négative rend les sujets plus susceptibles de poursuivre des stratégies plus simples dans les tâches de prise de décision (Wichary et al 2015). Bien entendu, cela signifie qu'un niveau d'excitation élevé peut parfois conduire à des comportements plus faibles, plus rapides et plus automatiques de manière non optimale. Une étude récente illustre ce danger en montrant comment l'excitation augmente la probabilité que les sujets soient la proie de publicités frauduleuses (Kircanski 2017). L'image qui se dégage est celle de sentiments métacognitifs - sous la forme de modèles descriptifs-directifs de la dynamique d'erreur attendue - guidant l'agent à travers des processus cognitifs rapides et lents répartis à diverses échelles spatio-temporelles dans un riche écosystème cognitif. Les sentiments métacognitifs contribuent à la cohésion des pratiques culturelles émergentes par l'harmonisation des dynamiques couplées, guidée par la modulation, et ils contribuent au désengagement de l'agent des pratiques culturelles en provoquant un changement entre les différentes politiques d'action (ou un changement progressif, dépendant largement des niveaux d'éveil). J'avais soutenu au chapitre 6 que les sentiments métacognitifs évaluent et régulent la cognition afin de diriger l'incertitude vers les niveaux optimaux de la MFP au fil du temps, et nous avons vu dans la section précédente que les pratiques culturelles ont tendance à réduire l'entropie au fil du temps. Les sentiments métacognitifs jouent également un rôle important ici, car ils guident le sujet pour qu'il s'engage dans les pratiques qui sont estimées les plus propices à la MFP au fil du temps. Il s'agira pour la plupart de pratiques culturelles négentropiques qui tendent à accroître la prévisibilité à court terme. Certaines d'entre elles seront des pratiques entropiques de type apprentissage et perturbateur. Les pratiques d'apprentissage auront tendance à susciter une valence positive à mesure que le sujet s'améliore dans la réduction de l'erreur dans un nouveau domaine. Les pratiques perturbatrices sont, là encore, les plus difficiles à résoudre. Une rapide réflexion phénoménologique met en évidence le sentiment d'anticipation et de perspicacité imminente de telles pratiques et, à la suite de l'AIT, on peut supposer que cette phénoménologie correspond à la promesse d'une montée en flèche imminente du succès des prédictions, qui pousse le sujet à s'engager joyeusement dans l'incertitude environnante. C'est un sujet sur lequel nous reviendrons lorsque nous reviendrons sur le sentiment de manakan dans la culture Evenki au point 8.5. Avant cela, voyons comment l'AIT peut développer la caractérisation de la désorientation comme un sentiment métacognitif en une théorie à part entière du phénomène.
VII — LA DÉSORIENTATION REVISITÉE Prenons le compte de l'expérience affective qui a émergé dans la section précédente pour éclairer l'affirmation selon laquelle la désorientation est un sentiment métacognitif qui évalue et régule les processus de navigation active du sujet. L'objet particulier de la désorientation correspond aux processus de navigation active du sujet. L'objet formel correspond à un manque de confiance dans lesdits processus. Lorsque les processus de navigation active sont censés diminuer leurs performances et générer un taux négatif (c'est-à-dire une valence négative), la désorientation apparaît. Les politiques d'action qui en résultent dépendront du contexte et de l'éveil.
— 159 —
Une grande partie de notre discussion dans les chapitres 1 à 4 a tourné autour des distinctions en ligne/hors ligne et des distinctions allocentrées/égocentrées concernant les représentations spatiales, qui n'ont pas encore été prises en compte dans le cadre du PP. Un bon point de départ est de revisiter l'idée de la carte cognitive, que j'ai brièvement introduite dans le chapitre 1, où j'ai souligné comment le système de représentation spatiale semble fonctionner en mode de satisfaction des contraintes, avec différents réseaux de cellules dans les régions hippocampique et parahippocampique qui suivent les représentations topologiques (cellules limites), métriques (cellules grilles), angulaires (cellules directionnelles) et de localisation (cellules de lieu) disponibles. Malgré les progrès étonnants des neurosciences de la cognition spatiale, il reste de nombreux points faibles dans notre compréhension. Il s'agit notamment de la dépendance des champs de lieu à la topologie et aux objectifs (Pfeiffer et Foster, 2013), de l'hétérogénéité des phénomènes de remapping (Colgin, Moser et Moser 2008 ; Kubie, Levy, Fenton 2019), et des séquences hippocampiques générées en interne pendant le sommeil qui vont au-delà des trajectoires passées (Gupta et al., 2010). Une tendance récente qui fait des progrès considérables dans le traitement de certaines de ces questions est le compte rendu prédictif de l'activité des hippocampes. Par exemple, Stachenfeld et ses collègues proposent que les cellules de lieu encodent une représentation prédictive des états futurs compte tenu de l'état actuel, ce qui, selon eux, donne lieu à des cartes prédictives qui s'adaptent à la structure des tâches, ce qui permet de résoudre le problème des champs de lieu qui dépendent de la topologie et des politiques de comportement (Stachenfeld, Botvinick, Gershman, 2017). Une autre étude montre comment un cadre de l'hippocampe comme effectuant une inférence d'état caché peut offrir une théorie unifiée du remapping (Sanders, Wilson et Gershman, 2020). La plus pertinente de ces explications prédictives pour la portée de ce chapitre est une théorie informatique récente qui considère l'hippocampe comme un modèle hiérarchique génératif (Stoianov, Maisto et Pezzulo, 2020). Selon cette théorie, le modèle génératif organise des expériences séquentielles (trajectoires) dans des contextes spatiotemporels cohérents (cartes). Un élément important ici est que le replay génératif (l'échantillonnage hors ligne d'expériences fictives) donne lieu aux séquences hippocampiques générées en interne, ce qui fournit un modèle pour la génération interne de séquences pendant le sommeil. Les cartes qui en résultent sont biaisées, codant la probabilité des éléments à des endroits donnés de manière à signaler leur importance (par exemple, les emplacements des objectifs sont fortement pondérés). Le modèle générateur est mis à jour de manière bayésienne en utilisant les trajectoires actuelles afin d'inférer la configuration spatiale de l'environnement. Il en résulte que le modèle génératif hiérarchique de l'hippocampe "utilise à la fois des cartes et des séquences comme structures (ou biais inductifs) pour organiser les éléments de l'expérience" (p. 15, Stoianov, Maisto et Pezzulo 2020). La vision de l'hippocampe comme modèle hiérarchique génératif est pertinente ici parce qu'elle s'inscrit parfaitement dans le cadre du PP34 et parce qu'elle fournit un lien solide entre la représentation spatiale et l'expérience spatiale. L'étude de Stoianov, Maisto et Pezzulo offre un compte rendu informatique empiriquement cohérent de la représentation spatiale, dans lequel les prédictions à un niveau élevé de la hiérarchie (cartes) influencent les prédictions à des niveaux inférieurs (séquences et éléments d'expérience). Pour des raisons de simplicité, je ferai référence au niveau de la hiérarchie produisant les cartes comme L2 et au niveau de la hiérarchie correspondant aux éléments d'expérience comme L1. Ce point de vue rejoint les aspects PP de cette monographie avec la distinction entre représentation spatiale en ligne et hors ligne introduite dans le chapitre 1. Une représentation spatiale hors ligne est un modèle génératif découplé. La représentation spatiale en ligne, quant à elle, est la configuration spatiale de l'environnement déduite à partir de l'expérience antérieure et des données actuelles. En d'autres termes, c'est l'hypothèse la plus probable concernant la configuration spatiale et, par conséquent, celle qui peuplera la conscience. 34 L'article en question est le premier résultat de recherche d'un grand projet de l'ERC (ThinkAhead) dirigé par Pezzulo pour
conceptualiser la planification humaine (la navigation étant un sujet principal) dans le cadre du PP.
— 160 —
La mise à jour du modèle génératif correspond à l'intégration de cadres de référence spatiaux allocentriques (cartes de l'environnement à des niveaux élevés de la hiérarchie) et égocentriques (configuration spatiale d'éléments proches autour du sujet). Dans cette optique, nous avons enfin un lien entre une carte cognitive et l'horizon d'intentionnalité qui structure notre expérience de l'environnement. La désorientation se manifeste lorsqu'on s'attend à un taux négatif qui est supposé être causé par une déficience dans les processus de navigation actifs. Prenons un cas standard de désorientation, un tel sujet se perdant dans un centre commercial (par exemple le sujet du scénario 1 au chapitre 4). Des prévisions L2 inexactes entraînent une erreur croissante aux niveaux inférieurs (par exemple, il est prévu qu'après avoir tourné un coin, un couloir apparaîtra à gauche, mais le couloir apparaîtra à droite), le système déduit qu'il y a un problème dans les processus de navigation active et la désorientation apparaît. Si un sujet est perdu, il y a de bonnes raisons de s'attendre à un effet en cascade d'augmentation de l'erreur, ce qui signifie qu'un décalage entre la L2 et l'erreur actuelle pourrait conduire à un épisode de désorientation relativement fort (c'est-à-dire très excitant) et fort (c'est-à-dire de haute intensité), car le système s'attend à une augmentation de l'erreur à venir. Bien entendu, la cause de l'augmentation prévue du nombre d'erreurs tendra à dépendre du contexte et sera plus spécifique qu'une déficience des processus de navigation active, qui est un terme assez général. Si le système déduit que le problème est dû à une carte cognitive inexacte, les prédictions en L2 seront mises à jour pour tenir compte de l'erreur entrante, ce qui entraînera un changement dans l'hypothèse globale, et l'individu commencera à explorer et à apprendre l'environnement à nouveau (formant de nouvelles prédictions en L1 et L2). D'autre part, si le système déduit que le problème est un problème d'intégration allocentrico-égocentrique (c'est-à-dire un problème d'exploitation des prédictions en L2 pour générer des prédictions précises en L1), un effort soutenu sera fait pour réinterpréter l'environnement afin de déterminer comment il s'intègre dans la carte cognitive. C'est le cas de nombreux épisodes de retournement. L'individu croit être face à la Place de la République (résultat de l'intégration de L2 et L1), réalise être face à l'Hôtel de Ville, et après une poussée de désorientation (la désintégration de L2 et L1) l'environnement semble pivoter de 180º (une réintégration de L2 et L1). Il y a cependant de bonnes raisons de parler de processus de navigation active au sens large, plutôt que de se limiter à un seul processus de navigation (ou à un ensemble de processus), comme l'intégration allocentrique - égocentrique. Là encore, pensez à suivre les instructions d'un appareil GPS qui se met soudainement à faire des pépins. Cela entraînera une augmentation attendue de l'erreur, dont la cause présumée sera les processus de navigation actifs. Dans ce cas seulement, le problème n'est pas la précision des cartes générées en interne en L2, ni l'intégration de ces cartes avec les prédictions de niveau inférieur et l'Erreur ; le problème est le dysfonctionnement d'un artefact cognitif. Un cas limite pourrait être une carte, car il est difficile de dire si l'hippocampe générerait de nouveaux modèles de haut niveau basés sur la carte (en L2) ou si la carte serait directement la source pour générer les prédictions en L1, cette dernière étant plus parcimonieuse et rationalisée. Pour l'instant, il s'agit d'une question ouverte, qui pourrait éventuellement être résolue par des recherches empiriques. Bien que la question des représentations externes ne soit pas directement abordée par Stoianov, Maisto et Pezzulo, leur système permet la coordination de l'hippocampe et du mPFC pour l'apprentissage bidirectionnel : "Les systèmes corticaux peuvent transférer des connaissances structurées et liées à des règles aux systèmes hippocampiques, ce qui peut contribuer à biaiser leur contenu et à réorganiser les expériences selon les règles apprises. Une possibilité théorique qui reste à tester dans le cadre de recherches futures est de savoir si, pendant les périodes d'oscillations couplées, les structures hippocampiques et corticales rejouent simultanément leurs modèles générateurs respectifs et s'entraînent mutuellement. "(p. 14, Stoianov, Maisto et Pezzulo, 2020).
— 161 —
Cette possibilité théorique ouvre la porte à l'apprentissage des représentations externes par l'hippocampe à médiation corticale. Néanmoins, comme le souligne le scénario de guidage GPS, la désorientation apparaîtra également lorsque les processus de navigation active ne dépendent pas principalement des modèles hippocampiques (comme c'est le cas, par exemple, de ce que nous avons appelé la navigation M4 au chapitre 7). Il en va de même pour le suivi d'un guide ou d'un système de signalisation. Le fait de considérer l'hippocampe comme un modèle hiérarchique génératif fournit une manière élégante et PP-congeniale de conceptualiser le cas standard de la désorientation, mais il est important d'élargir notre champ d'application, car la désorientation évalue la navigation d'une manière flexible qui va au-delà des représentations de l'espace générées par l'hippocampe. Comme nous l'avons vu dans la section précédente, l'éveil jouera un rôle important pour déterminer si l'émergence d'un sentiment métacognitif (dans ce cas, la désorientation) entraîne une modulation des processus cognitifs (les processus de navigation active) ou un changement (l'abandon des processus cognitifs actuels). Ceteris paribus, les situations de forte excitation entraîneront des réponses "rapides" (par exemple exploration, marche aléatoire...) par opposition à des réponses "lentes" (par exemple délibération). Plus l'excitation (c'est-à-dire le changement de rythme attendu) est élevée (faible), plus le gain de précision associé aux niveaux inférieurs (élevés) est important et plus (moins) l'apparition de la désorientation risque de provoquer un changement. Les épisodes de désorientation à faible niveau d'excitation auront tendance à conduire à une délibération explicite et à une modulation des processus cognitifs actifs. Les épisodes de désorientation à forte excitation auront tendance à entraîner des réactions d'action plus rapides et de niveau inférieur et à faire rapidement s'effondrer l'hypothèse globale actuelle afin de trouver un nouveau modèle qui s'adapte au mieux aux apports de niveau inférieur. En raison de la nature directive-descriptive des modèles générateurs, l'effondrement prendra la forme d'un changement à la fois de la politique d'action et de la configuration spatiale inférée de l'environnement. Par exemple, lorsqu'un sujet est "retourné", un nouveau modèle intégré de l'environnement apparaîtra, dans lequel l'intégration égocentrique et allocentrique se déplacera de 180º. Et lorsqu'un sujet se perd dans le métro, il peut sortir son appareil GPS et s'engager dans un processus cognitif distribué alternatif. Cependant, s'il n'y a pas de modèle alternatif viable de l'environnement, aucune nouvelle configuration spatiale inférée stable n'apparaîtra. Il en résultera un comportement exploratoire (et souvent un comportement sous-optimal comme la marche aléatoire) et la transformation de l'expérience du sujet que nous avons d'abord explorée au chapitre 2.
VIII — LA PHÉNOMENOLOGIE DE LA DÉSORIENTATION Etablissons maintenant comment le taux négatif attendu, supposé être causé par une déficience dans les processus de navigation active, a pour conséquence que le sujet subit une réduction de son espace de possibilités dans un environnement peu familier encadré par un horizon d'expérience incertain. En d'autres termes, relions le récit de la désorientation du PP au récit phénoménologique de la désorientation présenté au chapitre 2. Bien que ce ne soit pas le cas standard de la désorientation, les sujets de Binet, qui ont fait l'objet d'un retournement, fournissent à nouveau un cas de référence pour lancer la discussion. Nous avions souligné les aspects suivants dans l'expérience du retournement :
• Les cadres de référence égocentriques n'en sont pas affectés ; • les cadres de référence allocentriques sont presque toujours désalignés de 180º ; • elle empêche souvent la reconnaissance d'environnements connus ;
À ce stade, il devrait être clair que ce qui tourne mal dans le scénario de retournement est l'intégration des cadres de référence allocentriques et égocentriques. Bien que l'intégration L2-L1 ne
— 162 —
soit pas abordée dans le compte rendu de Stoianov, Maisto et Pezzulo, il semble clair, selon le point de vue standard du PP, que cette intégration est le résultat d'un processus dynamique de mise à jour des erreurs, de sorte qu'au fur et à mesure que le sujet se déplace, apportant de nouveaux éléments en vue, les prédictions générées en L2 s'ajustent en conséquence. Et encore une fois, les cartes en L2 sont des structures (biais qua inductifs) qui organisent les éléments de l'expérience. Dans le scénario retourné, même si les prévisions de bas niveau concernant la relation spatiale de l'agent avec l'environnement proximal (c'est-à-dire l'auto-localisation minimale) sont exactes, les prévisions concernant la relation spatiale de l'agent avec l'environnement non proximal sont inexactes. Comme je l'ai expliqué au chapitre 1, le désalignement commun de 180º résulte du rôle des trois axes corporels dans l'intégration allocentrique - égocentrique. En termes de PP, le corps du sujet est utilisé pour déduire la position de l'agent par rapport aux modèles générateurs produits en L2. Dans le cas le plus courant, le sujet tourne à gauche au lieu de tourner à droite par erreur (ou vice versa), ce qui entraîne un désalignement de 180º des éléments spatiaux que l'on s'attend à rencontrer en chemin. Si le corps sert à déduire l'intégration L2-L1, cette intégration sert à son tour à déduire la nature des éléments rencontrés dans l'expérience du sujet, ce qui entraîne souvent l'incapacité de reconnaître des environnements connus pour les sujets qui se retournent. Notre sujet retourné souvent cité, qui croit se diriger vers la place de la République, met un certain temps à reconnaître l'Hôtel de Ville, car c'était inattendu. Cela montre le rôle des cartes de L2 dans l'anticipation de l'identité des objets rencontrés à L1. Cette anticipation est un élément important du rôle structurel de l'horizon de l'expérience. Dans le chapitre 2, j'ai conceptualisé l'analogie entre les illusions d'orientation et les illusions de figure-fond (introduites pour la première fois par Charles Olman dans le contexte des environnements de microgravité) en termes d'instabilité transmodale dans l'intégration égocentrique et allocentrique qui est similaire à l'instabilité perceptive des illusions de figure-fond. L'image élaborée qui se dégage de ce chapitre nous donne une histoire beaucoup plus détaillée. Ici, l'instabilité de la désorientation correspond à la diminution de la confiance dans le modèle génératif de la L2. Reportez-vous à l'explication PP de la rivalité binoculaire (présentée au chapitre 5) : l'explication la plus probable de l'entrée visuelle passe de la maison à l'hypothèse du visage, dont aucune n'est stable (car aucune des deux ne peut expliquer l'erreur provenant d'un des yeux). Parfois, il y a aussi des moments brefs et intenables de flou maison-visage juste avant un changement. Cela est similaire aux scénarios de retournement : L'Hôtel de Ville apparaît de manière inattendue, ce qui provoque une surcharge d'Erreur. Deux choses se produisent à la fois : l'attente d'une erreur de navigation entrante (entraînant une désorientation) et l'incapacité du modèle L2 actuel à expliquer l'erreur. Toute la configuration spatiale de l'environnement semble tourner autour du sujet pendant un moment (le modèle L2 actuel ne parvient pas à expliquer l'erreur), et, après avoir reconnu le bâtiment, l'environnement semble tourner de 180º autour du sujet (un nouveau modèle L2 émerge). Lentement, le sentiment de désorientation se dissipe, car au fur et à mesure que le nouveau modèle explique l'erreur entrante, l'attente d'un taux négatif se dissipe. Maintenant, nous pouvons également expliquer pourquoi l'instabilité en question est transmodale : le modèle pertinent se situe à un niveau élevé et ce n'est que lorsque les prévisions descendent en cascade dans la hiérarchie que des prévisions spécifiques à la modalité sont générées. En d'autres termes, le modèle génératif de la L2 n'est pas spécifique à une modalité. La plupart de ce que nous avons dit à propos du retournement peut être appliqué au cas standard de désorientation (c'est-à-dire à l'instabilité de l'intégration allocentrico-égocentrique) : le modèle en L2 ne parvient pas à expliquer l'erreur, ce qui entraîne une instabilité transmodale et un taux négatif attendu attribué aux processus de navigation actifs ; et lorsqu'un nouveau modèle explique l'erreur, le taux négatif attendu se dissipe. En termes neutres pour le PP, il y a un pic d'incertitude sur la configuration spatiale de l'environnement par rapport au sujet et cela entraîne une désorientation - un manque de confiance dans les processus de navigation actifs.
— 163 —
Dans le deuxième chapitre, j'ai abordé l'anxiété, l'impuissance, la confusion et la diminution associées à la désorientation, et j'ai regroupé ces aspects affectifs de l'expérience sous le terme de réduction de l'espace corporel, caractérisée par un sens diminué des possibilités. Cette diminution du sentiment de possibilité s'explique mieux par un appauvrissement du paysage financier du sujet. Rappelez-vous que notre sujet retourné sait qu'il peut entrer dans le bâtiment qui se trouve devant lui (Hôtel the Ville), mais pas qu'il peut y entrer lui-même. Le bâtiment ne lui est plus familier, et il ne se sent pas le bienvenu pour l'explorer. Il existe une pléthore de possibilités d'action qui dépendent d'une navigation réussie, allant de la capacité à entrer dans des bâtiments familiers à notre capacité à trouver le chemin du retour. L'émergence du sentiment métacognitif de désorientation transforme le paysage des moyens en attribuant une faible précision à ces moyens dépendant de la navigation. Cette situation, associée à une valence négative (et éventuellement à une forte excitation), entraîne l'anxiété, l'impuissance, la confusion et la diminution susmentionnées qui sont caractéristiques de la désorientation.
IX — MANAKAN À première vue, il pourrait sembler que l'ACI ne soit pas adapté pour rendre compte des épisodes de désorientation positivement valorisés qui ont été examinés au chapitre 3. Après tout, une valence positive correspond à un taux d'erreur attendu positif, et il est difficile de voir comment une défaillance des processus de navigation actifs pourrait entraîner une attente de diminution de l'erreur. L'échec d'un modèle prédictif semble en contradiction avec l'attente d'une augmentation du taux de réussite des prédictions. Ce cas de "désorientation joyeuse" est le dernier (et peut-être le plus difficile) à résoudre pour le présent compte rendu. Comme nous l'avons fait dans le chapitre 3, nous commencerons par examiner l'expérience de manakan parmi les Evenki, qui servira à éclairer la dynamique fonctionnelle du cas de "désorientation joyeuse". Au chapitre 3, nous avions précisé que l'expérience non prescrite de l'environnement pendant le manakan n'est rendue possible que parce que pour l'Evenki, il n'y a pas d'intégration continue des cadres de référence égocentrique et allocentrique pendant l'orientation, étant donné que l'intégration allocentrico-égocentrique surdétermine l'horizon d'expérience du sujet par l'attente et l'anticipation. En revanche, un système de navigation complexe (impliquant des réseaux de chemins et de vastes connaissances hydrologiques et toponymiques) permet à l'Evenki de s'adapter habilement à l'environnement et de s'ouvrir à ses possibilités. Nous avons vu que les prédictions en L2 (cartes) contraignent les prédictions en L1, et le résultat de cela dans l'horizon d'expérience du sujet. Le degré auquel L2 contraindra L1 dépend de la pondération de la précision. Dans un cas paradigmatique de navigation en connaissance de levé, les modèles en L2 sont affectés d'une haute précision par rapport à L1, de sorte que les erreurs qui ne s'ajustent pas à L2 seront plus susceptibles d'être ignorées. Pour les personnes qui s'appuient sur une maîtrise pratique comme les Evenki, les modèles en L2 se voient attribuer une faible précision, de sorte que les erreurs auront tendance à être résolues aux niveaux inférieurs de la hiérarchie sans s'appuyer sur les modèles35 en L2. En d'autres termes, les cartes cognitives ne joueront pas un rôle substantiel dans la structuration de l'expérience environnementale du sujet. En conséquence, la configuration spatiale de l'environnement distal sera sous-déterminée pour les individus "maîtrise 35 Une question qui dépasse le cadre de ce chapitre est de savoir dans quelle mesure ces modèles de L2 sont générés dans le
cas des Evenki, ou s'ils sont simplement inexistants. Il se peut qu'un tel emboîtement de modèles générateurs dépende de processus de développement spécifiques à la culture. D'autre part, comme les Evenki reposent sur la connaissance des itinéraires, il semble certain que certaines séquences (c'est-à-dire des chemins plutôt que des cartes) jouent un certain rôle dans la limitation des prévisions de la L1. Cependant, il est probable que cela se traduise par des trajets vécus comme des moyens d'aller à certains endroits et qu'il n'existe pas de modèle global de la façon dont ces séquences s'imbriquent les unes dans les autres.
— 164 —
pratique" par rapport aux individus "carte mentale". Le résultat est que les erreurs survenant aux niveaux inférieurs de la hiérarchie sont résolues en adaptant rapidement les prévisions à ces niveaux inférieurs. C'est ce qui donne lieu à l'adaptation habile de l'Evenki - il s'agit d'une adaptation rapide aux vicissitudes d'un environnement changeant. L'approche Evenki est certainement un modèle léger, une approche de type "bag-of-tricks". Elle s'appuie sur une variété d'heuristiques et sur des connaissances pratiques et incorporées, et non sur une représentation de l'environnement distal. Au fur et à mesure de leur progression, les Evenki rencontrent de nouveaux lieux et avec eux, de nouvelles possibilités d'action jusqu'alors inattendues. La perception de ces possibilités est une affaire d'incarnation, qui implique une marche légère, des chaussures fines à travers lesquelles on peut sentir le sol, des branches qui se cassent pour laisser une trace, etc. (pour un compte rendu détaillé, voir Safonova et Sántha 2013). (pour un compte-rendu détaillé, voir Safonova et Sántha 2013). C'est un peu comme si l'on utilisait la stratégie de l'AOC pour attraper la balle, dans laquelle une certaine pratique incarnée sert à exploiter l'heuristique sans avoir besoin d'un modèle détaillé de la trajectoire de la balle. Comme pour la stratégie AOC, la marche d'Evenki repose sur un modèle de pondération de précision flexible pour s'adapter rapidement à l'erreur de bas niveau sans avoir besoin de modèles complexes en haut de la hiérarchie. Les niveaux inférieurs fonctionnent à des échelles spatiales et temporelles plus petites. L'aspect spatial signifie qu'ils n'impliquent des prévisions que sur l'environnement proximal. L'aspect temporel signifie qu'ils impliquent une anticipation beaucoup plus fine. Les prévisions de bas niveau sont plus immédiates, à la fois dans l'espace et dans le temps. Par conséquent, la navigation de "maîtrise pratique" se traduit par une expérience plus immédiate, sans le degré d'attente et d'anticipation qu'implique la navigation de "carte mentale". La maîtrise pratique est une façon d'embrasser l'incertitude. Il est facile de voir comment ce style d'orientation débouche sur le sentiment de possibilité et d'excitation éprouvé lors de Manakan : une marche habile entraîne un déroulement continu de nouvelles affordances. Nous pouvons également voir comment cela se traduit par une valence positive dans les termes de l'ACI : la réalisation continue des affordances naissantes et l'attente correspondante de nouvelles affordances à remplir se traduit par une attente de prévisions précises (car les affordances se traduisent par des actions qui répondent à des prévisions proprioceptives) - en d'autres termes, le paysage des affordances naissantes correspond à une attente de taux positif. Examinons maintenant les joyeux cas de désorientation. Au chapitre 3, nous avions passé en revue la littérature existante sur le sujet et énuméré une série de caractéristiques associées à de tels épisodes :
• l'absence d'origine, de lieu et de destination, • n'ayant pas de but précis, • dépasser les modèles simplistes, • la suspension des moyens de déplacement habituels, • se laisser attirer par les attraits du terrain, • l'aspect indécidable de l'expérience spatiale, • établir un nouveau type de relation avec l'espace, • une expérience de liberté.
Tout ce qui précède résonne avec la marche habile d'Evenki et l'expérience de Manakan telle que nous venons de la caractériser, à l'exception de deux caractéristiques : la suspension des modes de déplacement habituels et l'établissement d'un nouveau type de relation avec l'espace. De plus, comme nous l'avons vu au chapitre 3, nous ne pouvons pas dire que les Evenki sont désorientés lorsqu'ils font l'expérience de Manakan. C'est précisément parce que pour les Evenki, la marche habile est leur façon habituelle de se déplacer, alors qu'on ne peut pas en dire autant, par exemple,
— 165 —
des Situationnistes qui entreprennent une dérive. La façon habituelle de se déplacer de la plupart des sujets subissant des épisodes de désorientation joyeuse est largement basée sur la connaissance de l'enquête. La désorientation conduit parfois ces sujets à adopter une nouvelle façon de se déplacer, qui s'apparente à une marche habile, et ils éprouvent alors un sentiment semblable à celui de Manakan. Ce changement est précisément ce que des pratiques telles que la dérive visent à obtenir. La dérive est une pratique culturelle clairement entropique : à court terme, elle augmente l'entropie (désordre, imprévisibilité) dans le but d'établir des pratiques culturelles alternatives, c'est-à-dire de nouveaux types de relations avec l'espace. Des épisodes joyeux de désorientation surviennent alors lorsque la désorientation entraîne la suspension des moyens de déplacement habituels et l'établissement d'un nouveau type de relation avec l'espace. En termes d'AIT, la désorientation se traduit par un réarrangement du schéma des pondérations de précision : d'un mode à modèle lourd à un mode dans lequel l'erreur est résolue aux niveaux inférieurs de la hiérarchie sans s'appuyer sur les modèles L2. C'est ce que l'on entend par "dépasser les modèles simplistes" (simplistes car en schématisme, les modèles peuvent être incroyablement complexes !), ce qui conduit à embrasser "l'aspect indécidable de l'expérience spatiale" (car l'environnement distal n'est pas représenté et reste sous-déterminé) et à "se laisser entraîner par les attraits du terrain" (la génération de nouvelles affordances grâce à ce mode de navigation plus souple). C'est l'expérience de la liberté qui se manifeste dans les épisodes joyeux de désorientation. Le processus par lequel des épisodes joyeux de désorientation se traduisent par une valence positive est analogue au processus par lequel une marche habile se traduit par une valence positive pour les Evenki : de nouvelles possibilités d'action émergent et sont ensuite réalisées par l'action, ce qui se traduit par une attente de Taux positif. Il y a donc un sens dans lequel nous ne pouvons pas dire que le sentiment de désorientation a une valence positive : ce qui résulte en valence positive n'est pas, à proprement parler, le défaut d'un processus de navigation active, mais l'émergence d'un processus de navigation alternatif. L'insuffisance d'un processus de navigation actif se traduira nécessairement par un taux négatif attendu, mais cette valence négative sera intégrée dans l'attente plus large d'un taux positif si un nouveau processus de navigation prometteur apparaît. Par conséquent, les scénarios joyeux de désorientation seront une affaire hybride, du point de vue de la valence. La désorientation se traduira initialement par une valence négative, mais à mesure qu'un processus de navigation alternatif apparaîtra, la polarité changera lentement. Parfois, cela peut même se traduire par des sentiments mitigés, Rate étant censé avoir des polarités différentes selon l'échelle de temps (par exemple, négatif dans la période initiale de désorientation et positif par la suite), ce qui fait que les épisodes de désorientation positive s'apparentent à une entrée dans une fête surprise, à un plaisir coupable ou à des souvenirs doux-amers. Il y a toujours une menace de perdre dans un sens et pour que la valence négative de la désorientation soit surmontée par la valence positive de la désorientation "joyeuse", la menace de perdre son chemin doit être mineure. C'est ce que résume bien la maxime de Ray Bradbury : "Être perdu". Comme c'est effrayant. Être perdu en toute sécurité. Comme c'est merveilleux" (p. 61, Bradbury 2017). Un autre aspect à prendre en compte ici est que pour pouvoir profiter de "l'absence de but précis" et de "l'absence d'origine, de lieu et de destination", le sujet doit vraiment n'avoir nulle part où aller et aucune affaire à traiter. Il ressort clairement du corpus de rapports que nous avons recueillis que l'anxiété des sujets désorientés est fortement liée à la fois au danger perçu
Je marchais dans des rues étroites, peu éclairées et avec quelques personnes seules la nuit. Je me sentais en danger et anxieux parce que personne n'était avec moi et que l'environnement ne m'était pas familier.
et à avoir une destination que l'on essaie d'atteindre J'étais perdu dans le Centre des Arts de NYUAD en essayant de retrouver ma classe. C'était mon premier jour de classe, et il y avait une signalisation minimale indiquant où se
— 166 —
trouvaient les salles de classe. J'étais très anxieux parce que je ne voulais pas être en retard à mon premier cours.
Ces deux facteurs sont en accord avec notre compte. Si la situation est perçue comme dangereuse, on s'attendra à un taux négatif. Si le sujet a un objectif actif (par exemple, se rendre en classe), le fait de ne pas atteindre cet objectif entraînera également un pic d'erreur, et l'épisode de désorientation entraînera une attente de taux négatif.
X — CONCLUSION Au cœur de ce chapitre, de cette monographie et de la théorie proposée se trouve l'affirmation selon laquelle la désorientation est un sentiment métacognitif qui évalue et régule les processus de navigation active du sujet. Lorsque les processus de navigation active sont censés diminuer leurs performances, la désorientation apparaît. La performance des processus de navigation active (l'objet particulier du sentiment) est alors modélisée par un objet formel (qui correspond à la diminution de la confiance dans lesdits processus) qui est à la fois un modèle descriptif et une politique d'action directive. Les actions résultant de la désorientation dépendront à la fois du contexte et des niveaux de valence et d'éveil. Les épisodes de désorientation à faible niveau d'excitation auront tendance à conduire à une délibération explicite et à une modulation des processus cognitifs actifs. Les épisodes de désorientation à forte excitation auront tendance à entraîner des réactions d'action plus rapides et un effondrement rapide du modèle actuel et de la politique d'action guidant la navigation, ce qui entraînera un passage entre des processus cognitifs distribués (par exemple, de la flânerie à la recherche active d'un chemin, ou du suivi religieux d'une carte à l'exploration de l'environnement immédiat). En ce qui concerne la phénoménologie de la désorientation, deux aspects de la théorie sont mis en avant. Premièrement, la désorientation affaiblit l'influence des modèles spatiaux sur le façonnage de l'expérience environnementale, de sorte que l'horizon de l'intentionnalité des sujets désorientés devient moins déterminé à la suite de la désorientation. Deuxièmement, dans les cas paradigmatiques et négatifs, l'émergence de la désorientation transforme le paysage des moyens du sujet en drainant (par une pondération de précision) les moyens dépendant de la navigation, ce qui conduit à l'anxiété, l'impuissance, la confusion et la diminution qui sont caractéristiques de la désorientation. Enfin, dans certains cas, ce réaménagement du paysage de l'offre se traduit par l'émergence de processus de navigation alternatifs qui apportent les aspects positifs de la désorientation. Ce qui est intéressant dans les cas de désorientation joyeuse et dans la marche habile de l'Evenki, c'est que nos présentes considérations éclairent des questions plus larges. L'idée que l'activité fondamentale du cerveau est la génération et la mise à jour de prédictions n'implique pas nécessairement une vision de la cognition basée sur un modèle. La dextérité de l'Evenki à trouver son chemin montre que la cognition (dans ce cas, la navigation) peut fonctionner à des niveaux inférieurs, à la lumière du modèle, et se comparer favorablement à des processus cognitifs plus élevés, à la lumière du modèle. En outre, le contraste entre la navigation "maîtrise pratique" et la navigation "carte mentale" montre comment nos modèles du monde limitent notre expérience de ce monde et comment des modes de cognition alternatifs entraînent des modes d'expérience alternatifs. Parfois, ce n'est qu'après que nos cartes se sont effondrées que l'environnement se révèle mûr pour les possibilités. C'est là que réside l'ambivalence de la désorientation. Perdre la maîtrise du monde est une affaire dangereuse, et nos
— 167 —
sentiments sont sages pour nous le dire. Et pourtant, sans l'abandon momentané des anciennes méthodes, de nouvelles méthodes ne verraient jamais le jour.
