Artifacts and Original Intent: A Cross-Cultural...

© Koninklijke Brill NV, Leiden, 2008 DOI: 10.1163/156770908X289189

Journal of Cognition and Culture 8 (2008) 1–22 www.brill.nl/jocc

Artifacts and Original Intent:A Cross-Cultural Perspective on the Design Stance

H. Clark Barretta,*Stephen Laurenceb

Eric Margolisc

a Department of Anthropology, Center for Behavior, Evolution, and Culture, and FPR-UCLA Centre for Culture, Brain, and Development, UCLA, 341 Haines Hall,

Box 951553, Los Angeles, CA, 90095-1553, USAb University of Sheffield, Sheffield, UK

c University of Wisconsin, Madison, WI, USA* Corresponding author, e-mail: [email protected]

AbstractHow do people decide what category an artifact belongs to? Previous studies have suggested that adults and, to some degree, children, categorize artifacts in accordance with the design stance, a categorization system which privileges the designer’s original intent in making categorization judgments. However, these studies have all been conducted in Western, technologically advanced societies, where artifacts are mass produced. In this study, we examined intuitions about artifact categorization among the Shuar, a hunter-horticulturalist society in the Amazon region of Ecuador. We used a forced-choice method similar to previous studies, but unlike these studies, our scenarios involved artifacts that would be familiar to the Shuar. We also incorporated a community condition to examine the possible eff ect of community consensus on how artifacts are categorized. Th e same scenarios were presented to university student participants in the UK. Across populations and conditions, participants tended to categorize artifacts in terms of a creator’s intent as opposed to a diff ering current use. Th is lends support to the view that the design stance may be a universal feature of human cognition. However, we conclude with some thoughts on the limitations of the present method for studying artifact concepts.

KeywordsAnthropology, psychology, philosophy, concepts, cross-cultural analysis, artifact cognition, categorization, design stance

How do people decide whether something is a chair or a clock or a boat? And are such judgments dependent on diff erent principles than decisions about whether something is a frog or a tiger or a diamond? Many psychologists now think that while artifact concepts need to be distinguished from natural kind

2 H. C. Barrett et al. / Journal of Cognition and Culture 8 (2008) 1–22

concepts, the two have an important affinity. Both are said to exemplify aspects of essentialist thinking.

At the most general level, psychological essentialism maintains that ordinary categorization involves a fair amount of causal-explanatory reasoning and is not simply a matter of registering an item’s perceptual properties (Gelman, 2003). Psychological essentialism emphasizes that just because something looks like a skunk does not mean that it is judged to be a skunk. What matters is whether the item is thought to have the essence that is appropriate to skunks. Suppose, for example, that a creature started out as a cat and that scientists altered its appearance via plastic surgery, changing the shape of its legs, adding a white stripe down its back, installing a gland that secretes a noxious odor, etc. In the end, the creature would resemble a skunk far more than it would resemble a cat, but many people would maintain that it is still a cat (Keil, 1989). Th e thinking that underlies this judgment is that the creature’s skunk-like appear-ance lacks the right causal basis; it lacks the essence presumed to be shared by genuine skunks.

On standard treatments of psychological essentialism, ordinary people’s understanding of essences can be very sketchy. People may not be in a position to say much more about the skunk essence than that it is a hidden, internal property of these animals and that the essence determines, or strongly con-strains, the characteristic perceptual properties of skunks. To the extent that people have general ideas about what types of properties constitute an essence, the ones that are often cited are micro-structural properties – genes in the case of living kinds, chemical composition in the case of many non-living kinds.

At first glance, artifact concepts do not seem to follow the same general pat-tern as natural kind concepts. Chairs are not thought to have some hidden internal property that makes them chairs, and no one supposes there is a chair gene or chemical essence of chairhood. Also, though we can easily imagine how science might investigate skunks and force us to substantially revise our views about them, it is hard to see how the same thing could be true of chairs. However, proponents of psychological essentialism do not claim that natural kind concepts and artifact concepts implicate the same types of essences. Arti-facts are supposed to have their own distinctive type. As Paul Bloom has put it (Bloom, 2000, p. 163):

It is true that we don’t think of artifacts as having internal essences in the sense of natural kinds. But this doesn’t refute the view that artifacts are seen as having essences in the sense of having deeper causal properties that explain their super-ficial features and determine the categories they belong to. Instead of being biological and chemical, the essences of artifacts are social and psychological.

H. C. Barrett et al. / Journal of Cognition and Culture 8 (2008) 1–22 3

If there is a standard view about what this is, it is that the essence of an artifact has to do with the intention of the person who made it, the artifact’s creator or designer.

Th ere would be little interest in extending psychological essentialism to arti-fact concepts if a designer’s intent were just one property among others, where people decide that something is a chair, say, by simply summing up these prop-erties. Th e point of referring to a designer’s intent as the essence of an artifact is that the decision about whether something is a chair is an explanatory process similar to what goes on with natural kinds. Th e presumption is that people judge that something is a chair by inferring that the best explanation of its salient observable properties is that it has the right origin, that it was made with the distinctive intent associated with the category chair. Following current prac-tice, we will refer to this approach to artifact concepts as the design stance.

Proponents of the design stance often argue for it by drawing attention to its intuitive appeal. For example, Paul Bloom (1996) notes that typical ordi-nary artifacts cannot be categorized solely in terms of their appearance because their perceptual properties allow for too much variation. Not all clocks have a circular face with numerals imprinted along the edge. Th ere are also digital clocks, clocks shaped like footballs, and clocks for the blind that indicate the time non-visually. So, when we judge that something is a clock, we cannot rely on the distinctive appearance of a clock; there is no distinctive appearance. Bloom also notes that we cannot settle the matter by appeal to what an artifact does, for instance, the fact that clocks tell the time. After all, a broken clock is still a clock even though it does not tell the time; also, something may func-tion to tell the time (e.g., a shadow on a tree) and yet that does not make it a clock. Th e design stance does not face these complications because it does not take either appearance or function to directly establish whether something is a member of a given artifact category. Appearance and function merely enter into an explanatory inference as evidence about the category.

Apart from intuitive arguments, there is also some empirical evidence on behalf of the design stance, though much of this focuses on issues of cognitive development. For instance, Matan and Carey (2001) argue that the design stance begins to coalesce around the age of six. When an artifact’s original intended function is pitted against a diff ering current use, six-year-olds side with original intent, while younger children are indiff erent between the two. In contrast, Gelman and Bloom (2000) have found that even three-year-olds are sensitive to intent in naming artifacts. If told about a newspaper that becomes folded into the shape of a hat, they are more likely to name it using an artifact term (“hat”) than a material term (“paper”) when the transforma-tion is described as being intentional rather than accidental.


Developmental studies along these lines promise to off er important insight into the origins of the categorization strategies that underlie artifact concepts. At the same time, the focus on development is somewhat premature in that it overestimates the extent to which we understand the adult competence against which children are to be measured. We see at least four limitations of this research.

First, there are a number of related theories of artifact categorization that are conflated under the rubric of the design stance. Before turning to children, we need to tease these apart and determine which, if any, of these properly describes the pattern of adult categorization. We assume, at a minimum, that all proponents of the design stance take the categorization of artifacts to involve at least implicit causal-explanatory reasoning. Yet the exact role of the intent of a designer is controversial. Matan and Carey (2001) assume that what matters is ultimately the original intended function of an artifact, that is, the function that its creator had in mind. On this version, the essence of a chair might be taken to be that its creator intended it to be sat upon by a single individual. Bloom (1996) argues against this view, however, noting that such an intention would not distinguish chairs from similar yet distinct arti-facts (e.g., stools) and also that it is perfectly coherent that someone could make a chair without intending it to ever be sat on (e.g., as a work of art for display in a museum). For Bloom, the proper characterization of the design stance is still in terms of a designer’s intent, but the intention that matters is that the item was made to be a chair, not that it was made to serve a particular function.

If the diff erence between these two approaches were not enough, it is worth mentioning that many discussions of the design stance associate the view with the philosopher Daniel Dennett (1971). But while Dennett can be cred-ited with coining the phrase “the design stance”, and while Dennett also views the categorization of artifacts as, in some sense, an explanatory process, Den-nett himself argues that the intention of a designer does not have a special status in determining what category an artifact belongs to. For Dennett, the designer has no more say in the matter than an artist has in dictating the proper interpretation of his artwork (Dennett 1990). As he puts it, “the inven-tor is not the final arbiter of what an artifact is, or is for; the users decide that. Th e inventor is just another user . . . if others can find better uses for it, his intentions, clearheaded or muddled, are of mere historical interest” (1990, 186). Clearly, until we carefully distinguish these various approaches to arti-fact categorization and establish how adults actually think about artifacts, we are not going to be in a position to ask when and how children develop the design stance.


Consider, for example, the question of whether an object of one kind – say a teapot – can come to be an object of a diff erent kind simply in virtue of the intentions of its later users. In a study seeking to test the design stance, Adee Matan and Susan Carey sought to show that the original designer’s intent trumps other possible uses of the object. Matan and Carey presented brief stories about artifacts that are created by one person for a given purpose and then used by another for a diff erent purpose. In one of their stories, a woman makes something to water flowers with but another woman ends up using it to make tea. Th e question subjects are then asked is whether it is a watering can or a teapot. Matan and Carey found that adults side with original intent over current use about 80% of the time and have taken this to be solid evi-dence that people conceptualize artifacts just as the design stance says.

Intuitive evidence suggests, however, that sometimes current use trumps original intent. Bloom (2004) notes that Warhol and Duchamp created works of art from ordinary objects in apparently just this way, and that the same seems to at least sometimes be true of objects outside of art – for example, a house can become a church or a shop, and cease to be a house, simply in virtue of the intentions of its users (pp. 85-86). Such cases highlight the fact that standard presentations of the design stance tend to incorporate an ambiguity that needs to be resolved. Exactly whose intentions count and under what circumstances? Is it the original designer’s intention that matters, or can a later user “redesign” an object merely by imposing their own intentions upon it? What exactly is it to be the designer of an artifact? Notice that if having inten-tions about the use or kind of an object is enough, as Bloom’s house/church example suggests, then the design stance would make exactly the opposite pre-diction than Matan and Carey suppose. For present purposes we will put aside these complications and assume with Matan and Carey that the design stance is committed to the original designer’s intent as the factor that determines kind membership. We note, however, that there are important and unresolved issues about what exactly the design stance says about the matter.

Second, the empirical studies in support of the design stance typically pit design considerations against mere appearance or an alternative possible func-tion. Such studies may establish that designer’s intent has some role to play in categorization, but they are not enough to show that the essence of artifact concepts is given by the design stance, or even that artifact concepts have essences. To really put the design stance to the test requires an extensive set of contrasts showing how the explanatory interest of a designer’s intent relates to other considerations. Would designer’s intent still be so important if a large amount of time passed between the original intent and the current use? Or if the item were poorly suited to fulfilling the original intended function but


much better suited to the new function? Or if it were it were never used for the original function but often or habitually used for the new function? Or if it were used for the new function not just by one person, but by an entire com-munity? Or if an entire community came to regard the item as an object as having a diff erent primary function than its original function? Although some of these factors have been explored in related experimental paradigms, for example, in important studies by Kelemen (1999) and German and John-son (2002), there is a great deal of work left to be done. As things stand, it would be hasty to conclude that designer’s intent has the privileged role that advocates of the design stance suppose; it may be just one factor among others that guide the categorization of artifacts (Malt and Johnson 1998, Sloman and Malt 2003).

A related concern with experiments that are cited on behalf of the design stance is their reliance on stories that may seem somewhat artificial or short on details. In Matan and Carey’s study, there was a strong preference for saying that an object was of the category associated with the original intended func-tion. However, their subjects may have had difficulty supposing that the vari-ous objects underwent the change of use that the stories mentioned (e.g., that a watering can would plausibly be used as a tea pot). Or if the new uses were accepted, subjects might have been influenced by whether they thought an item was more likely to be a better at one or the other function. Th e fact that Matan and Carey provided so little information about the objects – nothing apart from their original intended function and the diff ering current use – left the subjects to make their own assumptions about these things.

A third potential limitation of these sorts of studies is that the question that is often asked involves a forced choice (Malt and Sloman, 2007a). Is it a water-ing can or is it a teapot? When adults side with the category corresponding to the original intent of the designer, the response is supposed to be indicative of the design stance. However, it ought to be acknowledged that the question is not an especially natural one. Th e worry then is that the answer that subjects give may not tap into a fundamental psychological operation – one that under-lies artifact categorization – but may tell us more about how people interpret the motivation behind the question (for more on this possibility, see Malt and Sloman (2007b,c) and Bloom (2007)). And if the forced choice is awkward for adults, it may be even more awkward for children, making it all the more difficult to draw conclusions about how children conceptualize artifacts.

Fourth, part of the motivation for looking at the emergence of the design stance in development is the supposition that it may be an innate, species-specific feature of the human mind. Th ough it is hard to find theorists who are perfectly explicit about this matter, we have little doubt that many who are


attracted to the design stance are motivated by the speculation that humans are proficient with artifacts partly because of our unique way of conceptual-izing them – i.e., in terms of a designer’s intent. Further, the connection between the abundance of artifacts in human life and the design stance is strengthened by the observation that non-human animals do poorly on theory of mind tasks and so are likely to have an impoverished understanding of intention. Perhaps it is their inability to conceptualize artifacts as we do that explains why their use of artifacts is so limited. It would be hasty, however, to suppose that current studies do establish that humans as a species conceptual-ize artifacts in terms of a designer’s intent. Th ese studies focus on a fairly nar-row adult population, viz., American college students and other high SES adults who live in technologically advanced Western societies. Even if the data establish that such people think of artifacts in terms of a designer’s intent, this would hardly show that the design stance is a human trait. Clearly, we need to look at other populations and the patterns of inference that are found in diverse cultural contexts. Th e study of natural kind concepts off ers a model for this kind of research. Recently, there has been considerable interest in the cross-cultural examination of concepts for living kinds, and the results have indi-cated a strong universal component to these concepts (e.g., Berlin, 1992; Medin and Atran, 2004). Similar work needs to be done on artifact concepts so that we can begin to identify which if any aspects of these concepts are universal.

Although no single study could possibly address all the concerns we have raised, the present study is meant to address some of these concerns and to encourage further investigation of how adults conceptualize artifacts. Here we adapt the basic methodology employed by proponents of the design stance but with two significant modifications.

One is that we perform a direct comparison between Western adults and adults in a technologically sparse culture, i.e., the Shuar of Ecuador. A previ-ous study with the Shuar revealed their tendency for functional fixedness, which in turn indicates that Shuar adults may associate a given artifact with a single prominent function (German and Barrett, 2005).1 However, that study did not directly solicit intuitions about naming or categorizing. By asking Shuar adults to consider scenarios that pit a designer’s intent against a diff er-ing use, we hope to determine whether the Shuar exhibit the same pattern of categorization as is claimed for Western adults.

Th e other modification is that our stories incorporate socially relevant information that might encourage categorization at odds with a designer’s

1 Functional fixedness occurs when the priming of a conventional use for an artifact makes it difficult to envision task-relevant atypical uses of the artifact.


intent. In particular, we were interested in whether the categorization of arti-facts may be sensitive to community norms regarding the use of an artifact. Community norms will typically line up with designer’s intent and, like intended function, may diff er from other possible uses of an item. But com-munity norms can also be pitted against designer’s intent. For example, if everyone in the community thinks that a given string-mesh item is used for resting upon (i.e., as a hammock), then even if the original designer intended it to be used for fishing with, people might think that it is a hammock, not a fishing net. On the other hand, if the design stance correctly describes how people conceptualize artifacts, then adding information about community norms should not shift people’s judgments away from a designer’s intent; so long as it is perfectly clear that the designer made it for fishing, the item should count as a fishing net, regardless of how it is currently understood. Another advantage of mentioning community norms is that it would not be surprising if these were given diff erent weight in diff erent cultures. For exam-ple, one might suppose that community norms matter less in a technologically sparse community where there may be more pressure to extract the maximum utility from the few available artifacts.

Methods

Stimulus Materials

We probed participants’ intuitions about artifact categorization using a forced-choice scenario method similar to that used in previous experiments (e.g., Matan and Carey, 2001). In this paradigm, participants are given verbal or written vignettes that describe the history of an artifact from its creation to its current use. In the present study, each vignette consisted of two critical parts. Th e first part described who created the artifact and for what purpose it was created. Th e second part described the causal scenario via which the artifact came to change hands and the circumstances under which its new owner used it. We employed two types of scenario, which we call the standard condition and the community condition. In the standard condition, which parallels that used by previous studies, the new owner finds the object and uses it for a pur-pose diff erent than that intended by its creator. Th e community conditions were identical with respect to both how the object reached its new owner and how it was used, but diff ered in that they described the new use as a matter of cultural consensus in the community of the new owner. All vignettes were fol-lowed by an “A or B” target question: “If you had to choose, would you say that the object is an A or a B?” (e.g., a fishing net or a hammock).


For example, here is the full text of a standard condition vignette as given to UK subjects:

An indigenous South American man made an object out of string for catching fish in the river. One day, when he was fishing with the object, it got washed away in the river and another man found it. Th e other man took it out of the river and hung it between two trees. He used the thing to rest in. From that day on, he always used it for resting.

Th e corresponding community condition was as follows:

Imagine two indigenous South American groups, the Ache and the Hiwi. Th e Ache make things with string. Th ey use these things to fish in the river. However, the Hiwi don’t do this. Like the Ache, the Hiwi use string to weave objects, but they don’t use them to fish. Instead, the Hiwi hang the objects between two trees, and use them to rest in.

A man from an Ache village made an object out of string to catch fish in the river, as the Ache do. One day, the man took the thing to the river to fish. When he was fishing with it, the river carried it away and another man, a man from a Hiwi village, found it. Th e Hiwi man took it out of the river and hung it between two trees. He used the thing to rest in, as the Hiwi do. From that day on, the object remained in the Hiwi village, and they used it like that.

Both types of vignette were followed by a forced choice, “A or B”, target ques-tion, in this case, “If you had to choose, would you say that the object is a fishing net or a hammock?”

All subjects received four diff erent scenarios, under the following experimen-tal constraints. Th ere were four diff erent story types, which are summarized in Table 1: net/hammock, basket/cage, pounder/marker and drum/bucket. Each subject received exactly one version of each of these stories.

Table 1Descriptions of artifact scenarios

Scenario name Description A Description B Target question

Net/hammock Object made of string, for catch-ing fish in the river

Object made of string, for hang-ing between two trees and resting

If you had to choose, would you say that the object is a fishing net or a ham-mock?


Basket/cage Object made of vines with hole in the end, where chickens can sleep at night

Object made of vines with hole in the end, for gath-ering vegetables

If you had to choose, would you say that the object is a basket or a chicken cage?

Pounder/marker Object made of wood, used to crush a starchy root to make the local drink called chicha

Object made of wood, for indicat-ing border of property

If you had to choose, would you say that the object is a chicha pounder or a property marker?

Drum/bucket Object made of wood with hole on one end, to hit with a stick to make rhythmic noises

Object made of wood with hole on one end, to keep rice in

If you had to choose, would you say that the object is a drum or a rice bucket?

Th e versions varied in two ways. One was the experimental manipulation, standard vs. community. Each subject received exactly two standard versions and two community versions. Which stories were matched with which exper-imental manipulation varied across subjects in a counterbalanced fashion, so that all story/manipulation combinations were presented. Second, order in which the artifacts were presented varied. For example, in the net/hammock story, some subjects were told that the original object was a net (A→B) and others that it was a hammock (B→A). Th e target question for this story, however, always asked about the net first. A→B/B→A orders were counter-balanced, to control for order eff ects. Although previous studies have counter-balanced the order of the items chronologically (such that the original category is sometimes presented first and sometimes second), we did not do this in the present study.2

Table 1 (cont.)

Scenario name Description A Description B Target question

2 We should note that previous studies have not found a statistically significant eff ect for the order in which the events of a vignette are told, i.e., whether the creator’s intent is mentioned first or whether the current use is mentioned first (Matan and Carey, 2001). A pilot study by Laurence, Margolis, and Siegal also found no such order eff ect with UK subjects.


Finally, the order in which the stories were presented to each subject was randomized, to control for this kind of order eff ect.

Th e stories were presented to Shuar and UK participants, as described below. Th e stories were constructed so as to be plausible to the Shuar. All of the artifacts are ones that could be, and sometimes are, made by the Shuar. We used the exact same artifacts in the scenarios given to UK subjects, though the stories were contextualized by explaining that they concerned indigenous South American people. Th us, for all artifacts used, it was highly plausible that they could be made by hand, by a single individual.

Th e community manipulation was also designed for maximum plausibility by describing an ethnic context with which the Shuar would be familiar. Th e Shuar are aware of other ethnic groups that have diff erent artifact traditions. For example, the culturally related neighboring group, the Achuar, manufac-ture blowguns by hand, whereas the Shuar do not. Although the Ache and Hiwi are actual indigenous South American groups, the Shuar are not familiar with them. Th e names were used to add further plausibility to the scenarios.

Participants

Shuar participants. Th e Shuar are an indigenous hunter-horticulturalist society in the Amazon region of Ecuador. Th e Shuar live in small villages of up to 300 peo-ple, with varying degrees of integration into mainstream Ecuadorian society. In all Shuar communities, many artifacts are still made by hand, including houses, fur-niture, jewelry, baskets, and other kinds of tools and household items.

Participants included 24 Shuar adults from three Shuar villages: 16 men (mean age 33.8 years, range 17 to 63) and 8 women (mean age 30.4 years, range 22–46).

UK participants. UK subjects included a total of 108 adult students at the University of Sheffield (Sheffield, UK). Two subjects declined to identify their sex. Of the remaining 106, there were 55 men (mean age 19.3 years, range 18–23) and 51 women (mean age 19.0 years, range 18–20).

Procedure

Shuar participants. Shuar participants were interviewed in their homes by the first author or a research assistant. Interviews were conducted in Spanish. All participants were Spanish/Shuar bilingual. Texts were read aloud from a sheet of paper, and subject responses were recorded in a field notebook.

UK participants. UK participants were given texts in the form of a question-naire booklet, with one story per page. Questionnaires were filled out at the


beginning of a class period. Subjects were not paid for participation. Texts given to UK subjects were identical to those given to Shuar subjects, except for brief phrases framing the South American cultural context for the UK students.

Texts of story scenarios are included in the Appendix.

Results

Data Coding

Each subject provided judgments for four stories. Each judgment was coded as “1” if the subject chose the original artifact category, and “0” if the subject chose the new artifact category. From these, each subject’s responses were summed and divided by the total number of responses given by that subject. Th is provided an index of the proportion of each subject’s responses that favored the original, as opposed to the new, artifact category. We used a pro-portion measure, rather than summing responses, to correct for missing data.

Th e only missing data occurred in a single story scenario, the basket/cage scenario, with Shuar participants. In this scenario, the Community condition (order B→A), had to be discarded because of a mistake in the text. In addi-tion, for UK participants in the drum/bucket scenario, Community condition (order B→A), the order was mistakenly given as A→B. Th erefore, all UK participants in the drum/bucket scenario, Community condition are treated as order A→B, but the data are not treated as missing.

Th e 1/0 coding system was used for all of the following analyses.

Is there an Overall Preference in each Population for Original Artifact Category?

For each subject, responses to all four stories were summed and divided by total number of responses given, as described above. Th ese indices were then compared against a chance proportion of 0.5 using a one-sample t-test.

Both populations showed a significant preference for the original artifact category, across conditions. Shuar participants preferred the original category 73.6% of the time (SD = 23.3%). Th is global preference was significantly above chance (one-sample t-test, two-tailed, t (23) = 4.97, P<0.001). UK par-ticipants preferred the original category 62.0% of the time (SD = 34.3%). Th is global preference was significantly above chance (one-sample t-test, two-tailed, t (107) = 3.65, P<0.001).


Was there a Diff erence between Shuar and UK Participants in Overall Preference for Original Artifact Category?

Values of the indices described above, proportion of responses preferring orig-inal artifact name, were compared between populations using a Mann–Whit-ney U-test (which does not assume equal variance or sample sizes between populations). Overall population preferences for original artifact category were not significantly diff erent (Z = − 1.27, P = 0.203).

Did the Standard/Community Manipulation have an Eff ect?

For each participant, an index was constructed for the participant’s two “stan-dard” story version responses, and for the participant’s two “community” story version responses. Mean values of these indices by population and condition are shown in Fig. 1.

For Shuar participants, overall preference for original artifact category was 75.0% in standard conditions (SD = 25.5%), and 70.8% in community con-ditions (SD = 35.9%). Both of these were significantly above chance in the

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Shuar UK

Population

Mea

n pr

efer

ence

for

orig

inal

art

ifact

cat

egor

y (+

/- 1

SEM

)

Standard

Community

Figure 1. Mean preference for original artifact category by population and condition (standard versus community).


direction of preference for original artifact category (one-sample t-tests, two-tailed; standard: t(23) = 4.80, P<0.001, community: t(23) = 2.85, P<0.01). However, neither was significantly diff erent from each other (paired sample t-test, two-tailed: t(23) = 0.526, P = 0.604).

For UK participants, overall preference for original artifact category was 64.4% in standard conditions (SD = 38.1%) and 59.7% in community con-ditions (SD = 41.3%). Both of these were significantly above chance in the direction of preference for original artifact category (one-sample t-tests, two tailed; standard: t(107) = 3.92, P<0.001, community: t(107) = 2.45, p<0.05). However, neither was significantly diff erent from each other (paired sample t-test, two-tailed: t(107) = 1.20, P = 0.234).

Th us, while the diff erence between standard and community conditions was in the direction expected if community norms shift subjects’ judgments away from original intended function, in neither population was it significantly so. Both populations showed a significant preference for original artifact cat-egory, irrespective of condition.

Was there an Eff ect of which Artifact Category was Presented First?

We did not counterbalance artifact presentation with respect to chronological order of events in the story (i.e., original artifact category first or second). However, we did counterbalance the actual artifact category labels. For exam-ple, in half of the net/hammock stories, the net was the original category (A) and the hammock the new category (B), and in the other half, the order was reversed (B first, A second). In contrast, the order that these items were asked in the target question were always the same, namely, “A or B?” (“net or ham-mock?”). Th us, there was the possibility for order eff ects if, for example, sub-jects’ preferences were aff ected by order of presentation in the stories and in the target question.

To assess this, indices were computed for all cases in which item A was pre-sented first, and B second (A→B), and for all cases in which B was presented first and A second (B→A).

Among Shuar participants, preference for original artifact category was 81.3% (SD = 28.8%) in A→B conditions, and 66.7% in B→A conditions (SD = 35.1%). Th is diff erence was not significant (paired sample t-test, two-tailed, t(23) = 1.57, P = 0.129).

Among UK participants, preference for original artifact category was 69.3% (SD = 37.3%) in A→B conditions and 51.0% in B→A conditions (SD = 45.3%). Th is diff erence was significant (paired sample t-test, two-tailed, t(107) = 4.35, P<0.001).


Th us, in both populations, judgments in favor of original artifact category were higher in conditions in which the order of artifact categories presented was A→B, as opposed to B→A. However, this diff erence was only significant for UK participants.

Were there Eff ects of Order of Story Presentation?

We randomized the order in which stories were presented to participants. To examine eff ects of story order, we conducted a one-way repeated measures ANOVA for each participant population, with order in which the story appeared as the single factor, and category judgment (0 or 1) as the dependent variable.

For Shuar participants, there was no eff ect of story order on preference for original artifact category (F(3,16) = 0.423, P = 0.739). For UK participants, on the other hand, there was a significant eff ect of story order on preference for original artifact category (F(3,105) = 37.7, P<0.001), such that preferences for original artifact category increased from the first story presented to the last story presented.

Discussion

Overall our data lend further support to the design stance approach to artifact categorization. Subjects in both cultures tended to side with the category cor-responding to the creator’s original intent. Although this tendency was stron-ger for the Shuar than for UK subjects, the diff erence was not statistically significant. Also, the standard versus community manipulation had no signifi-cant eff ect.

In some ways, both of these results are surprising. Th e Shuar live in a society that off ers little exposure to advanced technology and where most familiar artifacts are made by hand. Also, given the paucity of artifacts in Shuar society, there is an obvious need to employ whatever tools happen to be available so long as it is minimally eff ective. College students in the UK, in contrast, are surrounded by mass-produced artifacts that permeate every aspect of their lives, and the prevalence of highly-specialized artifacts ensures that there is a tool for every job. Th ese diff erences are so striking that one might well expect to see a corresponding diff erence in how artifacts are conceptualized. Intui-tively, the Shuar have good reason not to hold any special allegiance to the original intent of an artifact’s creator. Doing so makes it harder to approach matters with the flexibility that is appropriate to recruiting the few available


artifacts in novel circumstances. But our results indicate, on the contrary, that there is a pattern of categorization that the Shuar share with Western subjects. When given a choice between original intent and current use, both tend to side with original intent and do so regardless of contravening community norms. To this extent, our results replicate previous findings on behalf of the design stance (Matan and Carey, 2001) and are consistent with the specula-tion that the design stance may be a universal feature of human cognition.

It is important, however, to recognize the limitations of the present study. One is that while our procedure controlled for various potential order eff ects, we did not vary the presentation of the chronological order of events in a vignette. Th is means that subjects always heard about the original intent first and the subsequent use second. Th e concern is that a primacy eff ect, where people find earlier material to be more salient, might favor answers in the same direction as the design stance. We think that this is unlikely, however, since earlier pilot studies with UK subjects using diff erent materials showed no sta-tistically significant order eff ect of this kind (see note 2).

A second limitation to the current study is that there are many factors that are potentially relevant to real-world categorization decisions that are omitted in our vignettes. Earlier we mentioned that previous studies that are taken to support the design stance generally leave out most information about an arti-fact apart from what it its creator intended and how it is currently used. We noted that this is a somewhat artificial arrangement, since real-world categori-zation decisions may be sensitive to a wide variety of information that is nor-mally available and considered to be relevant. Our own study aims to remedy the situation by reintroducing one potentially relevant piece of information, viz., the community norm relating to a given artifact. We also made an eff ort to make the transformations of the artifacts plausible (e.g., one could plausibly imagine a hammock being used as a net and vice versa). However, we’d be the first to admit that these improvements are only a beginning. We found that the community condition on its own did not have a statistically significant eff ect with either population, but perhaps presenting and systematically vary-ing other variables would have a statistically significant eff ect (either on their own or in combination with one another). As we noted earlier, it is not clear whether designer’s intent would trump current use if there were a significant temporal gap between creation of the artifact and its current use, or if an item were relatively poorly suited to fulfilling its original intended function as com-pared to its current function, and so on. Clearly, much work still needs to be done along these lines in order to fully evaluate the design stance.

One striking feature of our data for Western subjects is that we found a weaker reliance on original intent than Matan and Carey found. While 80%


of Matan and Carey’s adults favored original intent, our UK subjects sided with original intent only 62% of the time. Th ere are a number of possible explanations for this. One is that what we have here is a modest cultural dif-ference (a diff erence between two Western industrial societies). Another pos-sibility is that our larger subject pool (108 adult subjects as compared to Matan and Carey’s 16 adult subjects) is a more accurate reflection of at least the West-ern tendencies. Another possibility is that these data are only an incomplete gloss on a more complex underlying pattern. A significant subpopulation of subjects in our study consistently chose on the basis of the community norm over the original function (see Fig. 2). Among our Shuar subjects, 7 individu-als out of 24, or 29% of subjects, chose original intended function in half the scenarios or fewer. Among our UK subjects, 51 out of 108, or 47%, chose original intended function in half the scenarios or fewer. For comparison, 3 out of 16, or 19%, of adult participants in Experiment 1 in Matan and Carey (2001) chose original intended function in half the scenarios or fewer.

Now there are two ways of interpreting the consistent minority pattern. One is to say that they are essentially noisy data. Th is explanation would seem very reasonable to us if 80–90% of subjects, across a large sample size, consistently

0%

10%

20%

30%

40%

50%

60%

70%

80%

Half or fewer of scenarios More than half of scenarios

Proportion of scenarios in which subjects choseoriginal intended function

Per

cent

of s

ubje

cts

Shuar

UK

Figure 2. Percent of scenarios in which subjects chose original intended function.


chose in accordance with the original function. But given that only 56% of our subjects (pooled) chose in accordance with original function more than half the time, it seems only reasonable to consider alternative accounts of the data. One such possibility is that there are multiple stable cognitive styles in evidence here. Some subjects may conceptualize artifacts in accordance with the design stance while others do not. Perhaps certain people regularly default to the community’s norms governing an artifact’s current use.3 Even though we found the same overall trend in favor of the design stance in a technologi-cally sparse hunter-horticulturalist society as in a technologically advanced Western society, this may not mean that human beings as a species conceptu-alize artifacts in accordance with the design stance.

Finally, an even more radical possibility is that our results are strongly col-ored by the forced choice method that we have adopted (in common with other researchers in this area), where subjects have to decide between two artifact categories. Subjects may have only a relatively small tendency toward the design stance, which is amplified by the forced choice task. In the end, the forced choice method may not be the best way of determining how people conceptualize artifacts. While it is certainly encouraging to see that the methodology can be adapted for use in studying adults in significantly diff er-ent cultures, and while the method also has the advantage that it can be employed with children, there is still reason to wonder whether it accurately reveals people’s categorization dispositions. As we noted earlier, the question Is it an A or is it a B? may elicit unintended pragmatic inferences about the questioner’s intent. Rather than directly addressing the choice between A and B, subjects may find themselves reasoning about what such an odd question could mean. If this is a concern for Western subjects, it is all the more so when we turn to cultures that have less experience with social science surveys and context-free questionnaires.4 And if subjects do feel that the question is

3 Anecdotal evidence suggests that something similar may hold for natural kind concepts. Putnam (1975) famously introduced a thought experiment about a substance that looks and tastes like ordinary water but has a very diff erent chemical composition, XYZ instead of H2O. Putnam’s intuition was that it is a forgone conclusion that the substance is not water even though it has water’s superficial properties. But our experience teaching the thought experiment in the classroom is that students are often split between those who share Putnam’s intuition and those who do not. Perhaps people who have the opposing intuition conceptualize natural kinds diff erently than Putnam. It certainly seems as if they do.

4 A related concern is that the forced-choice method may not be sensitive to the conditions in which an evolved system for thinking about artifacts would have its natural application. Such a system, if one exists, is unlikely to have developed under pressure to identify definitive categories. It’s more likely that the adaptive function for such a system would have been to organize manmade items for a given purpose, for forming inductive inferences about their functionality, their mechanical organization, or their social significance (or some combination of these).


unnatural, perhaps feeling that there is no answer to the question of what the item “really” is, then this would further undermine the design stance.

Conclusion

Th is study off ers further experimental support for the design stance, in the form of the first cross-cultural study of this hypothesis. Contrary to our initial expectations, our data from the technologically sparse Shuar society were broadly consistent with previous findings in support of the design stance. Nonetheless, we believe caution is in order before inferring that the design stance is part of a universal human conceptual system for representing arti-facts. Here we have outlined what we believe are some of the remaining open questions about the design stance, possible theoretical alternatives to it, and key variables for future work to test. It seems clear that how people think about articrafts is an important part of the story of human evolution, and we suspect that there is much yet to discover about it. We hope that the ideas and data presented here will help to further refi ne hypotheses about the nature of human articraft cognition.

Acknowledgements

Th anks to the FPR-UCLA center for Culture, Brain, and Development for funding, to the villages of Chinimpi, Putuntsa and Shauk for their participa-tion in this study, and to Nicole Wilson for help collecting data in the fi eld.

References

Berlin, B. (1992). Ethnobiological Classification: Principles of Categorization of Plants and Animals in Traditional Societies. Princeton University Press, Princeton, NJ.

Bloom, P. (1996). Intention, History, and Artifact Concepts. Cognition 60, 1-29.—— (2000). How Children Learn the Meanings of Words. MIT Press, Cambridge, MA.—— (2004). Descartes’ Baby: How the Science of Child Development Explains What Makes Us

Human. William Heinemann, London.—— (2007). More Th an Words: A Reply to Malt and Sloman, Cognition 105, 649-655.Dennett, D. (1971). Intentional Systems. Journal of Philosophy 68, 87-106.—— (1990). Th e Interpretation of Texts, People, and Other Artifacts. Philosophy and Phenom-

enological Research 50, 177-194.Gelman, S. (2003). Th e Essential Child: Origins of Essentialism in Everyday Th ought. Oxford

University Press, New York, NY.Gelman, S. and Bloom, P. (2000). Young Children Are Sensitive to How an Object was Created

when Deciding What to Name It. Cognition 76, 91-103.German, T. and Barrett, H.C. (2005). Functional Fixedness in a Technologically Sparse Culture.

Psychological Science 16, 1-5.


German, T. and Johnson, S. (2002). Function and the origins of the design stance. Journal of Cognition and Development 3, 279-300.

Keil, F. (1989). Concepts, Kinds, and Cognitive Development. MIT Press, Cambridge, MA.Kelemen, D. (1999). Th e scope of teleological thinking in preschool children. Cognition 70,

241-272.Malt, B. and Johnson, E. (1998). Artifact Category Membership and the Intentional-Historical

Th eory. Cognition 66, 79-85.Malt, B. and Sloman, S. (2007a). Artifact Categorization: Th e Good, Th e Bad, and the Ugly.

In E. Margolis and S. Laurence (Eds.), Creations of the Mind: Th eories of Artifacts and Th eir Representations. Oxford University Press, Oxford.

—— (2007b). Category Essence or Essentially Pragmatic? Creator’s Intention in Naming and What’s Really What. Cognition 105, 615-648.

—— (2007c). More than Words But Still Not Categorization. Cognition 105, 656-657.Matan, A. and Carey, S. (2001). Developmental Changes within the Core of Artifact Concepts.

Cognition 78, 1-26.Medin, D. and Atran, S. (2004). Th e Native Mind: Biological Categorization and Reasoning in

Development and across Cultures. Psychological Review 111, 960-983.Sloman, S. and Malt, B. (2003). Artifacts are not ascribed Essences, nor are they treated as

Belonging to Kinds. Language and Cognitive Processes 18, 563-582.

Appendix A: Story Texts, UK Versions

Note: Th e Shuar received Spanish translations of these texts, which were identical, except for contextualizing information, such as “an indigenous South American man”.

Net/Hammock

Standard A→B

An indigenous South American man made an object out of string for catching fish in the river. One day, when he was fishing with the object, it got washed away in the river and another man found it. Th e other man took it out of the river and hung it between two trees. He used the thing to rest in. From that day on, he always used it for resting.

If you had to choose, would you say that the object is a fishing net or a hammock?

Community A→B

Imagine two indigenous South American groups, the Ache and the Hiwi. Th e Ache make things with string. Th ey use these things to fish in the river. However, the Hiwi don’t do this. Like the Ache, the Hiwi use string to weave objects, but they don’t use them to fish. Instead, the Hiwi hang the objects between two trees, and use them to rest in.

A man from an Ache village made an object out of string to catch fish in the river, as the Ache do. One day, the man took the thing to the river to fish. When he was fishing with it, the river carried it away and another man, a man from a Hiwi village, found it. Th e Hiwi man took it out of the river and hung it between two trees. He used the thing to rest in, as the Hiwi do. From that day on, the object remained in the Hiwi village, and they used it like that.

If you had to choose, would you say that the object is a fishing net or a hammock?


Basket/Cage

Standard A→B

An indigenous South American man used some vines to make an object with a big hole in one end that his family could use to bring vegetables back from their garden. One day, another man found the object and took it to his house. Th is man used the object as a little house where his chickens could sleep at night. From that day on, he used the object this way.


Community A→B

Imagine two indigenous South American groups, the Ache and the Hiwi. Th e Ache use vines to make objects that have a big hole in one end. Th e Ache use these objects to bring vegetables back from their gardens. However, the Hiwi don’t do this. Like the Ache, the Hiwi use vines to make objects that have a big hole in one end, but they don’t use these to bring vegetables back from the garden. Instead, the Hiwi use them as little houses where their chickens can sleep at night.

A man from an Ache village used some vines to make an object with a big hole in one end to use to bring vegetables back from the garden, as the Ache do. One day another man, a man from a Hiwi village, found it. Th e Hiwi man took it and brought it to his village. He used it as a little house where his chickens could sleep at night, as the Hiwi do. From that day on, the object remained in the Hiwi village, and they used it like that.


Chicha Pounder/Property Marker

Standard A→B

An indigenous South American woman made an object out of wood that she used to crush a starchy root to make the local drink called chicha. One day, another woman found the object and took it to her house. She was having an argument with her neighbor over where the boundaries of her property were, so she put the object in the ground to indicate the border of her property. From that day on, she always used it like that.


Community A→B

Imagine two indigenous South American groups, the Ache and the Hiwi. Th e Ache make objects out of wood that they use to crush a starchy root to make the local drink called chicha. However, the Hiwi don’t do this. Like the Ache, the Hiwi make objects out of wood, but they don’t use them to make chicha. Instead, the Hiwi use them to put in the ground to indicate the borders of their properties.

A woman from an Ache village made an object out of wood to prepare chicha, as the Ache do. One day another woman, a woman from a Hiwi village, found the object and took it home. She put it in the ground to indicate the border of her property, as the Hiwi do. From that day on, the object remained in the Hiwi village, and they used it like that.



Drum/Rice Bucket

Standard A→B

An indigenous South American man made an object out of wood that had a hole on one end. He hit the object with a stick to make rhythmic noises. One day, another man found the object and took it to his house. He used it to keep his rice in. From that day on, he used it that way.


Community A→B

Imagine two indigenous South American groups, the Ache and the Hiwi. Th e Ache make objects out of wood that have holes in one end, and they hit these objects with sticks to make rhythmic noises. However, the Hiwi do not do this. Like the Ache, the Hiwi make objects out of wood that have a hole in one end, but they do not use them to make rhythmic noises. Instead, the Hiwi use the objects to keep their rice in.

A man from an Ache village made an object out of wood that had a hole in one end, and he hit it with a stick to make rhythmic noises, as the Ache do. One day another man, a man from a Hiwi village, found the object and took it home. He used it to keep his rice in, as the Hiwi do. From that day on, the object remained in the Hiwi village, and they used it like that.


7

Research Challenges for Digital Archives of 3D CulturalHeritage Models

DAVID KOLLER, BERNARD FRISCHER, and GREG HUMPHREYS

University of Virginia

The increasing creation of 3D cultural heritage models has resulted in a need for the establishment of centralized digital archives.

We advocate open repositories of scientifically authenticated 3D models based on the example of traditional scholarly journals,

with standard mechanisms for preservation, peer review, publication, updating, and dissemination of the 3D models. However,

fully realizing this vision will require addressing a number of related research challenges.

In this article, we first give a brief background of the virtual heritage discipline, and characterize the need for centralized 3D

archives, including a preliminary needs assessment survey of virtual heritage practitioners. Then we describe several existing

3D cultural heritage repositories, and enumerate a number of technical research challenges that should be addressed to realize

an ideal archive. These challenges include digital rights management for the 3D models, clear depiction of uncertainty in 3D

reconstructions, version control for 3D models, effective metadata structures, long-term preservation, interoperability, and 3D

searching. Other concerns are provision for the application of computational analysis tools, and the organizational structure of

a peer-reviewed 3D model archive.

Categories and Subject Descriptors: H.5.1 [Information Interfaces and Presentation]: Multimedia Information Systems

General Terms: Documentation, Management, Standardization

Additional Key Words and Phrases: 3D models, cultural heritage

ACM Reference Format:Koller, D., Frischer, B., and Humphreys, G. 2009. Research challenges for digital archives of 3D cultural heritage models. ACM

J. Comput. Cult. Herit. 2, 3, Article 7 (December 2009), 17 pages.

DOI = 10.1145/1658346.1658347 http://doi.acm.org/10.1145/1658346.1658347

1. INTRODUCTION

Technological advances in our ability to measure the physical world and in our computer modeling capa-bilities have led to the creation of an increasing number of high-fidelity 3D models of cultural heritageobjects and environments. These models are often supplemented by additional metadata informationas well as synthetic reconstructions of missing data. This growing collection of models brings with itnew needs and opportunities in the scholarly community and the public it serves: efficient access to,interoperability between, and scientific authentication of the 3D models. In addition, there is a pressingneed for archival methods to ensure survivability of these important cultural artifacts.

The authors gratefully acknowledge support from the National Science Foundation (grants 0535118, 0713295).

Authors’ addresses: D. Koller (corresponding author), B. Frischer, and G. Humphreys, Department of Classics, University of

Virginia, Charlottesville, VA 22904-4204; email: [email protected].

Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided

that copies are not made or distributed for profit or commercial advantage and that copies show this notice on the first page

or initial screen of a display along with the full citation. Copyrights for components of this work owned by others than ACM

must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers, to redistribute to

lists, or to use any component of this work in other works requires prior specific permission and/or a fee. Permissions may be

requested from Publications Dept., ACM, Inc., 2 Penn Plaza, Suite 701, New York, NY 10121-0701 USA, fax +1 (212) 869-0481,

or [email protected].

c© 2009 ACM 1556-4673/2009/12-ART7 $10.00 DOI 10.1145/1658346.1658347 http://doi.acm.org/10.1145/1658346.1658347

ACM Journal on Computing and Cultural Heritage, Vol. 2, No. 3, Article 7, Publication date: December 2009.

7:2 • D. Koller et al.

We advocate the creation of open repositories of scientifically authenticated 3D cultural heritagemodels, with standard mechanisms for preservation, peer review, publication, updating, and dissemi-nation of the 3D models. In this article, we consider a number of technical research issues that mustbe addressed to realize this goal of effective 3D archives.

We first give a brief background of the virtual heritage discipline. Then we characterize the need forcentralized 3D archives, including a preliminary needs assessment survey of virtual heritage practition-ers. This is followed with a description of several existing state-of-the-art 3D cultural heritage repos-itories. Finally, we enumerate the primary technical research challenges that must still be addressedto realize an ideal 3D archive, including issues of digital rights management, clear communication ofuncertainty and associated metadata, versioning, preservation, and interoperability.

1.1 3D Virtual Cultural Heritage

3D cultural heritage, or “Virtual Heritage,” is the relatively new branch of knowledge that utilizesinformation technology to capture or represent the data studied by archaeologists and historians of artand architecture. These data include three-dimensional objects such as pottery, furniture, works of art,buildings, and even entire villages, cities, and cultural landscapes. There are three ways that digitaltools support this kind of data representation.

(1) Existing objects can be automatically measured with 3D capture technologies such as laser scanningor photogrammetry.

(2) Damaged or no-longer extant objects can be modeled by hand with 3D software (e.g., AutoCad,Maya, 3D Studio Max, MultiGen Creator).

(3) Because file formats of almost all commercially available software can be easily translated, we cancombine models made by capture and by hand to create hybrids.

The primary uses of these representations include:

—cataloguing and documentation;

—public outreach and education;

—historical studies; and

—experimental architectural and urban history.

From this list, it can be seen that virtual heritage is a continuation of the traditional activities whichscholars and scientists active in the area of cultural heritage studies have been pursuing for severalcenturies. Whereas the cultural heritage community previously used static 2D forms of documenta-tion (plans, sections, elevations, reconstructions) created on paper and published in print, it is nowincreasingly using 3D interactive digital tools. This transformation of the medium of expression andpublication has meant that virtual heritage has spread rapidly through a large, well-established field,which has generally embraced the new technologies in recognition of their obvious superiority to whatthey have replaced.

1.2 History of Virtual Heritage

In the 1985 Proceedings of the Annual Conference of Computer and Quantitative Methods in Archae-ology (abbreviated as CAA), we find the first major article on 3D capture: Leo Biek’s [1985] article onstereovideo. The first article on 3D hand-modeling was published four years later in the 1989 CAAProceedings by Arnold et al. [1989], who discussed computer models of the temple precinct at RomanBath and the Roman military bathhouse at Caerleon, South Wales. In the early 1990’s, J. D. Wilcock’sprediction from the first CAA conference in 1973 [Buckland and Wilcock 1973] that computers would


Research Challenges for Digital Archives of 3D Cultural Heritage Models • 7:3

be used to reconstruct archaeological sites started to be realized. A well-known overview of the vir-tual heritage scene was given in the 1997 book by M. Forte and A. Siliotti [1997] Virtual Archaeol-ogy, in which they described several dozen computer reconstructions of sites around the world. For amore recent history of this aspect of the virtual heritage field, see Frischer et al. [2002] and Frischer[2008].

From this 1997 book, we can see that the early computer models served the purpose of illustration, andpublications tended to focus on methods and technologies supporting the creation of such illustrations.It is noteworthy that almost all the computer models in Forte and Siliotti’s [1997] book were madeby private companies, and that none give authorship credit to a professional archaeologist. By thelate 1990’s, the situation had changed, and with the drop in the costs of creating computer models,many archaeologists started their own 3D projects. Many of these were presented at CAA 1998 inBarcelona [Barcelo et al. 2000], and every annual meeting since then has included an increasing numberof papers on this topic. Meanwhile, other professional organizations around the world have starteddevoting conference time to virtual heritage. These include the “VAST International Symposium onVirtual Reality, Archaeology and Cultural Heritage,” “Computer und Archaeologie” (sponsored by theCity of Vienna, Austria), “Virtual Retrospect” (presented every other year in Biarritz, France), and“The International Society for Virtual Systems and Multimedia.” As another positive sign that virtualheritage has now reached maturity and mainstream recognition, we note the fact that in 2007 theAssociation for Computing Machinery (ACM) is sponsoring publication of a new journal devoted to thefield, the Journal on Computing and Cultural Heritage.

There is no reliable count of the number of scholars active in this discipline, nor do we know thenumber of heritage sites modeled to date. An estimate by a leading practitioner in April 2006 put thenumber at well over 300 [Sanders 2006]. It is safe to predict that this number will continue to growgeometrically in the next decade as 3D modeling software and data capture systems are mastered andtheir price falls, even to the point where they are free. See, for example, Google’s recently introducedfree 3D modeling program, SketchUp, or, for free 3D data capture and processing, MeshLab and theEpoch 3D Webservice [Vergauwen and Gool 2006].

2. NEED FOR CENTRALIZED 3D CULTURAL HERITAGE ARCHIVES

The rapid spread of tools for virtual heritage has inevitably created some new problems. This articleoutlines a number of research directions to address these problems, and sets forth a major challengefacing the virtual heritage community: the lack of central repositories for the collection, peer review,publication, updating, preservation, and distribution of 3D models of cultural heritage data.

The lack of such repositories has had serious negative consequences, as has been recognized by thecommunity itself (see the following on the preliminary results of a recent needs assessment survey).Foremost among these is the lack of any plan for the preservation of virtual heritage models. Ironically,the very scientists who are doing so much to document and preserve the world’s heritage resourcesare doing little to conserve their own digital products. Secondly, there is no tool to discover whether aparticular cultural monument has been digitally captured or modeled; even a Google or Yahoo searchis generally fruitless, and there is no Web resource specifically in this area. Thirdly, the creators ofvirtual heritage products are not rewarded by their peers for creating digital models and related writtendocumentation, because there is no peer review agency evaluating the quality of the work and acceptingit for publication. Finally, the end-user, whether a scientist, student, or member of the general public,has no way to access a model, let alone an online resource offering a wide range of high-quality scientific3D digital models of cultural heritage monuments which he or she may wish to download.

It is our contention that the best way to address these problems is through creation of one or morecentral services, following the general models of a scholarly press and a photographic archive. The



Table I. Demographics of Respondents (note that provision of this information was optional)

Year of birth No. Sex No. Country No. Position No. Models Created No.

1932-42 1 M 25 Australia 1 Technician 1 1-10 11

1943-52 2 F 6 Canada 4 Cultural ministry 2 11-25 9

1953-62 6 France 4 Indep. scholar 4 26-50 4

1963-72 11 Germany 1 Assist. Prof. 11 51+ 5

1973-82 9 Italy 2 Assoc. Prof. 5

Mexico 1 Full Prof. 11

Romania 2 Other 11

S. Korea 2

Turkey 1

USA 6

former offers methods of quality control, recruitment of creators, and outreach to the academic com-munity; the latter, methods for economies of scale and sustainability through aggregation and revenuesharing between creators and distributors.

A long-term objective, then, should be the creation of centralized, open repositories of scientificallyauthenticated virtual environments of cultural heritage sites. By scientifically authenticated, we meanthat such archives should accession only 3D models that are clearly identified with authors with appro-priate professional qualifications, and whose underlying design documents and metadata are publishedalong with the model. Uncertainties in the 3D data and hypotheses in the reconstructions must beclearly documented and communicated to users. Thus, we envision these 3D archives as realizations ofa peer-reviewed journal of 3D cultural heritage models and associated scholarship.

2.1 Needs Assessment Survey Results

Beginning in 2006, we have been running a needs assessment survey conducted online and at pro-fessional meetings at which creators of 3D content present their work to their peers. The purposeis to determine whether this “virtual heritage” community feels the need of an online service thatwould peer review, publish, and provide long-term storage for 3D models of heritage sites. The surveywill be administered several more times in 2007 before being closed and the final results published.Thus far, we have received 54 responses from attendees at the annual meetings of “Computer Appli-cations and Quantitative Methods in Archaeology” (held in Fargo, ND, USA from April 18–21, 2006)and “Computer und Archaeologie 2006” (Vienna, Austria, October 18–20, 2006). We have also cre-ated an online version of the survey and have invited members of the Humanist Listserv to take it athttp://www.iath.virginia.edu/save/.

The survey respondents so far come from ten countries. Their median age is 38 years old. The majorityare males (80%) and professors at universities (60%). They have already created, on average, between11 and 25 models of cultural heritage sites, although a surprising number (17%) report that they havemade over 50 models. The demographics suggest that the activity of 3D modeling is starting to spreadacross the planet, with notable (and not surprising) concentrations in North America and Europe.Academics have adapted well to this new way of illustrating the monuments they have long studied.

The results strongly suggest that the virtual heritage community does think that the time for creatingsuch a service has come. Thus far, 90% of all respondents answered “definitely a need” or “probably aneed” to the question, “Is there a need for a central repository and distributor of 3D models?” Therewas also support for offering the following services along with publication: long-term preservation (94%called this “definitely” or “probably” important), peer review (80% called this “definitely” or “probably”important), and interoperability of models with each other (82% called this “definitely” or “probably”important). Brief highlights of the survey results to date are given in Table I and Table II.



Table II. Views About the Need for a Central Repository of 3D Models and Its

FeaturesNeed for central Long-term Interoper- File format

repository? preservation? Peer review? ability? conversion?

Definitely 23 28 20 15 8

Probably 26 19 20 26 20

No 1 2 7 4 12

No opinion 4 1 3 5 8

3. PRIOR 3D CULTURAL HERITAGE ARCHIVES

A number of 3D archives of cultural heritage models have been created in recent years. While none ofthem yet realize the full vision of a peer-reviewed, searchable, interoperable repository of scientificallyauthenticated 3D models, they represent the state of art and the first steps toward this goal. In thissection, we review a few such 3D cultural heritage archives.

3.1 Stanford Digital Michelangelo Project

The Digital Michelangelo Project at Stanford University created a 3D archive of Michelangelo’s sculp-ture by digitizing the original marble artworks in Florence, Italy [Levoy et al. 2000]. Ten statues,including the David, were digitized with high-resolution laser scanners, resulting in 3D models with0.25 mm precision consisting of hundreds of millions of polygons each. The models, as well as the rawrange scans and metadata, are distributed via an archive at the Project Web site. However, due to thecontract with the Italian authorities, the 3D models and raw data are distributed only to establishedscholars, who submit an application to the project director for review. An alternative means of access tothe models is provided for the general public, via a novel remote rendering system, “ScanView,” whichis designed to protect the 3D shape of the models from theft while still allowing interactive examinationof the models [Koller et al. 2004].

3.2 Stanford Digital Forma Urbis Romae Project

Another very large archive of scanned 3D models at Stanford University is the Digital Forma UrbisRomae Project [Koller et al. 2006]. The Forma Urbis Romae, a giant (18 by 13 meters) marble mapof Rome constructed in the early 3rd-century, is an archaeological artifact of primary topographicalimportance, as it depicts the ancient city at a scale of approximately 1:240. Over one thousand fragmentsof the map have been recovered; the 3D shape of these surviving fragments were digitized with a laserscanner by the Stanford team, with the intent of developing geometric matching algorithms to searchfor new joins among the fragments. The 3D models serve as the centerpiece of an online informationsystem about the fragments and the marble map, which also includes digital images, textual analysis,and linked bibliographical data for each fragment, thus providing a complete digital resource for bothscholars and the public. Again, access to download the full 3D geometry of the models is limited toapproved scholars, although the ScanView protected 3D graphics system is available for anyone tointeractively view the fragments.

3.3 UCLA Digital Roman Forum

A real-time virtual reality 3D reconstruction of the Roman Forum as it appeared it late antiquity wasdeveloped at UCLA beginning in 1997 [Frischer et al. 2006]. Models of the interiors and exteriors ofeach building and monument in the Forum were created using 3D modeling software and authenti-cated by a scientific committee of leading archaeological experts (Figure 1). The real-time 3D databasewas converted to a digital library format suitable for distribution on the Internet, including Quick-Time movies of the 3D models, a time-space map of the Forum, comprehensive metadata about each



Fig. 1. Digital model of the Roman Forum with detailed view of the Rostra, foreground, with Temple of Saturn and Temple of

Vespasian, background (left) and a shot of the Forum today from the same vantage point. Photo left is copyright of The Regents

of the University of California; photo right is copyright Bernard Frischer.

component of the 3D models, photographs of extant remains, and appropriate articles from Romantopography sources. This presentation of a 3D virtual heritage reconstruction is a pioneering exampleof an effective scientific committee organization for authentication, as well as an excellent example ofdetailed metadata describing the uncertainties and design decisions inherent in the hypothetical 3Dreconstructions. However, the online archive lacks a means for users to access the actual 3D modelsand to explore or analyze their geometry.

3.4 CyArk 3D Heritage Archive

The CyArk 3D Heritage Archive is a recent effort to create an online Internet-based repository forcultural heritage site data developed through laser scanning, digital modeling, and other spatial tech-nologies [CyArk ]. Currently the CyArk prototype features data from ten prominent archaeologicalsites, including Angkor, Pompeii, and Ancient Thebes. As the archive was founded by the developersof the Cyra Technologies laser scanner, many of the sites include 3D point cloud survey data, withaccompanying images and map data linked together in a spatially-navigable interface with search-able information about each site. The 3D data can be viewed either with a downloadable commercialpoint-based viewer plugin, or as a QuickTime video.

3.5 Other 3D Archives

A number of other archives of 3D virtual heritage are under construction and being deployed. TheDigital Archive Network for Anthropology and World Heritage (DANA-WH) at North Dakota StateUniversity is a distributed network of databases of anthropological objects, with custom client soft-ware for viewing and measuring the 3D models of fossils and artifacts [Clark et al. 2002]. The DigitalHammurabi Project at Johns Hopkins University aims to perform 3D scanning and archiving of largesets of cuneiform tablets [Digital Hammurabi Project ]. Organizations such as the Arts and HumanitiesData Service (AHDS) are promoting and maintaining repositories of digital cultural heritage resources,including 3D datasets. Large established digital libraries like the Perseus Project at Tufts Universityhave incorporated some 3D representations into their collections, both in image-based QuickTime for-mats as well as 3D geometric formats, such as the VRML collection of Greek vases in Perseus [Shiawet al. 2004]. Commercial archives for 3D models, such as the TurboSquid online clearinghouse for 3Ddigital artists [TurboSquid], often include cultural heritage models and address many of the issues ofsearchability and portability to different file formats.



4. RESEARCH CHALLENGES

There persists a number of unresolved research issues concerning the effective management and dis-semination of cultural heritage models in a curated 3D archive. Here we enumerate these challenges,in an approximate order of priority, reviewing related prior work and suggesting new directions forresearch.

4.1 Digital Rights Management for 3D Graphical Models

Three-dimensional models of classical artworks, archaeological artifacts, or historic sites are often veryvaluable intellectual property, representing treasured cultural heritage and patrimony. The curatorsof physical artifact collections may be reluctant to allow digitization of their holdings if they wouldcompletely lose control of the virtual representations. Likewise, many cultural heritage scholars andcontent developers would be unwilling to participate in a centralized digital archiving effort that doesnot offer some guarantees about the security and trustworthy dissemination of their intellectual prop-erty. Thus, providing for the digital rights management of 3D models is of paramount importance inthe development of 3D cultural heritage archives.

Disseminating interactive 3D content securely is a difficult problem, due to the variety of attacksallowed by the traditional computer graphics pipeline (these possible attacks are discussed in Kolleret al. [2004]. Most of the prior research in this area has focused on the a posteriori detection of piracy, notits a priori prevention. While piracy detection and the related problem of verifying model authenticity isimportant, prevention is the key to successful integration of 3D models into cultural heritage archives.Shortly, we suggest new research directions for protected dissemination of 3D models, both with aposteriori steganographic watermarking methods, and techniques for developing a trusted graphicspipeline to prevent piracy.

4.1.1 Watermarking 3D Models. Steganographic approaches to tracking and authenticating datainvolve hiding an embedded signature or “watermark” in the digital document to be protected. Spe-cialized watermarking techniques for 3D models can prove highly useful in cultural heritage archives.First, watermarking provides an easy way to verify that a model is authentic, as one can embed theunique signature of a trusted authority in the model itself; users can check that their version is nota digital forgery. Secondly, such a signature could be used to allow only certain similarly signed pro-grammable shaders to be executed by the graphics pipeline, allowing sophisticated use of moderngraphics hardware for visualization while preventing an entire class of model recovery attacks. In bothof these applications, fragile watermarks suffice; these are watermarks that can be easily removed. Thisis because removal of the watermark will either cause the model to lose its sign of authenticity, or tonot be visualized properly.

In other cases, it may be advantageous to distribute 3D models with a hidden robust watermarkembedded in it; such a watermark is difficult to remove without significantly degrading the qualityand usefulness of the model. The presence of a robust watermark on a 3D model would allow themodel owner to establish a chain of custody if the model was ever redistributed in violation of themodel’s licensing agreement, and pursue enforcement action if desired. For example, 3D models ofstatues downloaded from a 3D archive could be signed with a unique watermark identifying the initiallicensee.

Dozen of research papers in the last decade have described approaches for watermarking 3D graphicalmodels, with varying degrees of robustness to attacks that seek to disable watermarks through alter-ations to the 3D shape or data representation. Many of the most resilient 3D watermarking schemesare based on spread-spectrum frequency domain transformations, that embed watermarks at multiplescales by introducing controlled perturbations into the coordinates of the 3D model vertices [Praun



Fig. 2. The ScanView client-side low-resolution 3D rendering (left) is replaced by images from the server-side high-resolution

renderings (right) when the user stops moving the mouse.

et al. 1999; Ohbuchi et al. 2002]. Other appearance attributes of the model besides the 3D geometrycan be exploited to hide information as well [Zhang et al. 2002].

There is significant room for improvement in both fragile and robust watermarking of 3D modelsfor cultural heritage applications. In particular, the analysis of the impact of watermarking on thequality of the eventual 3D visualization has been at best ad hoc; typically it suffices to say “the modelslook indistinguishable.” Not only is this approach lacking in rigor, it also gives no insight into thesteganographic bandwidth: how much foreign information can be embedded in a model without affectingthe quality of the final rendering. Further research could formally study the perceptual impact ofwatermarking, investigate techniques to guarantee that a watermark will not be perceptible, and designnew watermarking schemes that minimally impact the perceived quality of the rendered image for anygiven amount of information to be embedded. The need to evaluate the perceptibility of the distortionsintroduced by the watermarking process is especially important for cultural heritage applications,where the human user may care about very subtle features in the model. For example, an art historianstudying patterns of chisel marks in the Digital Michelangelo Project 3D models would like to be certainthat the subtle variations they are studying represent the chisel mark, and not the watermark!

4.1.2 A Trusted 3D Graphics Pipeline. While steganographic approaches provide a possible meansfor authenticating 3D models and tracking piracy, the highly valuable and sensitive nature of somecultural heritage 3D models may require methods to prevent piracy a priori rather than merely detectit after the fact. We thus seek new methods for providing a “trusted 3D graphics pipeline” that canprotect the 3D data shared via an archive, while still allowing users to interact with and visualize the3D models.

The state of the art in providing protected interactive 3D graphics of cultural heritage objects wasrecently advanced by Koller et al. [2004] with the introduction of their “ScanView” remote renderingsystem. ScanView prevents recovery of a 3D model’s geometry by providing the user with only a verylow-resolution version of the 3D model for navigation purposes, and delegating high-quality renderingsto a physically inaccessible remote secured server that contains the high-resolution model. These high-resolution images are automatically requested by the client software on “mouse up” events and therendered images passed back from the server replace the corresponding low-resolution rendering (seeFigure 2). Slight distortions are introduced into the images returned by the server to defend againstany computer vision reconstruction attacks. Remote rendering provides an immediate level of securitybecause the user does not have any direct access to the 3D model itself, but only rendered 2D images.Additionally, a centralized rendering server as used in the ScanView system has the advantages ofconsolidating graphics resources, does not require clients to perform computationally-intensive 3D



rendering, and allows the 3D archive maintainer to fully log and monitor usage of the 3D models at avery fine-grained level.

However, such a remote rendering scheme for protected access to 3D models suffers a number ofdrawbacks. First, a rendering server with sufficient network bandwidth and low latency to all clientsmay not always be available. The networked nature prohibits offline usage, which may be important tousers such as archaeologists working in the field. Additionally, while the ScanView system is effectivefor viewing and examining static 3D models such as statues or individual artifacts, remote rendering isnot appropriate for more continuous, real-time visualization applications such as 3D walkthroughs orimmersive virtual reality viewing of large archaeological sites. Thus, we encourage investigation of newmethods, such as those outlined next, which may hold better promise for allowing protected sharing of3D cultural heritage models.

Secure graphics hardware. Digital rights management systems are designed to process their dataon potentially hostile hosts; they accomplish this through the use of trusted players. A trusted player isa device that can decrypt and manipulate data, but cannot be tampered with. Digital Video Disc (DVD)players operate in this manner; a manufacturer that wishes to sell a DVD player applies for a decryptionkey that is embedded in the player. This key can then be used to decrypt DVD videos for playback. Ina similar way, future 3D graphics cards for personal computers could have a hidden sequence of bitsembedded in the hardware that could be used to decrypt 3D models (they would be distributed inthe archive in encrypted form) and then render them. This capability would prevent attackers fromrecovering 3D models by intercepting the graphics API commands, or by capturing the model with ahomemade malicious graphics adapter. Some care would need to be taken to ensure that an attackercould not recover the 3D model using either straightforward readback mechanisms or through othersystematic misuse of the graphics API. Of course, the well-known fate of DVD encryption provides acautionary tale for future DRM efforts.

While this hardware systems approach is certainly sound from a security point of view, the archi-tectural implications of adding cryptographic capabilities to graphics hardware have not been studied.Such research is vital to demonstrating the practical viability of this technique for protecting 3D cul-tural heritage archives, because modern visualization systems rely heavily on the high performance ofthe graphics pipeline.

A related problem is securing of graphical shaders. Modern graphics hardware contains both pro-grammable vertex and fragment engines, and for secure graphics it would be necessary to ensure thata malicious user does not use a vertex program to manipulate or discover the encrypted 3D geometry.Research should investigate the possibility of adding support for signed shaders to the graphics pipeline,so that only shaders whose signatures match those of the model could be used. Cultural heritage 3Dmodels may make use of complex shaders to achieve sophisticated rendering effects, so it may alsobe important for content owners to be able to hide the contents of trusted shaders as well as the 3Dgeometry.

Ciphermodels and encrypted rendering. One interesting possible approach for allowing protecteddistribution and usage of sensitive 3D models involves the application encrypted computation tech-niques for digital rights management [Paris and Fiedler 2005]. Whereas more traditional computersecurity methods address the execution of untrusted software code on a trusted computer, encryptedcomputation techniques attempt to allow data to be processed on potentially hostile computers. Ifsuch a scheme could be implemented, the 3D models would be distributed in an encrypted form,along with a custom viewer that performs the critical rendering functionality directly on the en-crypted representation. The 3D model data would never be decrypted, and is thus safe from theft orabuse.



While the encrypted computation problem is widely believed to be unsolvable in general, it is perhapsplausible for the case of encrypted 3D rendering, as the projection of a 3D model onto a 2D screen is doneby a very specific, linear function which may be amenable to encrypted computation. The limitations ofencrypted computation are currently an open research problem, and encrypted rendering appears farfrom practical at this time, but warrants further investigation.

Defending against reconstruction attacks. Even if a model can be distributed from an archive andtransformed from its 3D description into a 2D image in a secure way, a malicious user can employ avariety of attacks that draw on computer vision algorithms to recover the model from a sequence ofhigh-quality renderings. Koller et al.’s ScanView system employs pseudorandom perturbations to makesuch an attack ineffective. These random errors need to be introduced into both the viewing parametersand the lighting parameters to defend against both geometric and radiometric attacks. In ScanView,the perturbations are applied when the user releases the mouse button; they rely on change blindnessto ensure that the user is not distracted by the introduced randomness. This technique is very effective,but is hard to apply directly to a system that is constantly rendering high-quality imagery. Furtherresearch should explore fully the extent to which perturbation defenses can be applied to high-qualityinteractive rendering, and also design more forward-looking defenses that do not require error to beintroduced at all.

Protected geometric analysis. Whereas the techniques described before are concerned primarily withallowing protected rendering and display of 3D models, an ideal secure dissemination system would alsoallow users a greater degree of geometric analysis of the protected 3D models without further exposingthe data to theft. For example, scholarly users of the ScanView system have expressed interest inmeasuring distances and plotting profiles of 3D objects for analytical purposes, beyond the simple 3Dviewing supported in the current system. Database researchers have developed privacy-preserving datamining techniques that extract useful knowledge from databases without significantly compromisingdata security [Verykios et al. 2004], and we are interested in the applicability of similar methods to theproblem of protecting 3D data.

4.2 Metadata for 3D Cultural Heritage Data

While geometric models of cultural heritage sites are the primary information structures to be archivedin the 3D repositories that we are addressing, the goal of scientific authentication and scholarly publi-cation requires that such models be supplemented with corresponding textual metadata. We considerthree types of metadata that authors will associate with the 3D models deposited in an archive [Frischer2003].

(1) Catalog metadata include the Dublin Core Metadata Initiative elements such as title, creator,subject, publisher, etc. These are useful for indexing, searching, and citation aids for users, andallow some level of interoperability with other digital library systems that leverage Dublin Coreconformity (or other similar standards).

(2) Commentary metadata include information about the evidence for the various elements of thereconstruction and commentary on the modeling design decisions. Alternative views in the scientificliterature pertaining to the model are noted, and the reasons for rejecting them are given.

(3) Bibliographical metadata include all the sources, published and unpublished, used in making amodel.

The inclusion of these metadata with 3D models are critical to achieve transparency to the user ina way analogous to notes, commentary, and bibliography in a traditional academic print publication.While it is unlikely that perfect reconstruction accuracy relative to an ancient monument’s original



appearance can be achieved for complex models, offering complete transparency about the underlyingsource data and decision making process determines the scientific authenticity of the resulting models.

As part of our research efforts into methods for managing and disseminating 3D cultural heritagemodels, we should investigate techniques for organizing associated metadata and its effective presen-tation to users. One particularly interesting concern is methods for allowing interactive exploration ofrelevant metadata displayed in corresponding locations in the 3D environment. Large urban modelsof ancient cities, for example, may contain thousands of individual buildings, each of which will poten-tially have distinct associated metadata. Thus the metadata will need to be spatially referenced to theunderlying 3D model, and the 3D exploration interface linked to an appropriate display methodologythat combines presentation of 3D and textual metadata elements. The deep inheritance trees that formthe basis of many cultural heritage metadata architectures are not likely to be appropriate for organiz-ing 3D metadata, and thus we suggest research into alternative models based on modular networks ofshallow metadata objects.

4.3 Uncertainty Visualization for 3D Models

The recent surge in usage of digital visualization tools has created an unfulfilled need for new datavisualization techniques which clearly communicate the uncertainty in the underlying data and thedepicted phenomena. Leaders in the scientific visualization community have called for researchers toaddress this representation of uncertainties in an effort to take visualization research to the next level[Johnson and Sanderson 2003]. Current scientific visualization researchers have proposed methodsincluding error bar glyphs, blurring, and false coloring of data to depict uncertainty in their datasets.Researchers in the areas of geography and GIS have described techniques specialized for geospatialdata uncertainty visualization, which is typically applied to a 2D domain [MacEachren et al. 2005; Pang2001]. At the recent NSF-sponsored “Summit on Digital Tools for the Humanities” (Charlottesville, VA,USA, September 28–30, 2005), a group of leading humanities scholars identified “Visualization of Time,Space, and Uncertainty” as one of four primary processes of humanistic scholarship especially ripe foradvancement. They called for a suite of software tools and visual conventions to be developed that allowrepresentation and visualization of degrees of uncertainty.

Three-dimensional virtual environments of cultural heritage sites are particularly well suited to mo-tivate and benefit from advances in uncertainty visualization techniques. Many 3D cultural heritagemodels are reconstructions of sites that are not fully extant nor documented, and thus the reconstruc-tions contain a large number and variety of hypothetical, uncertain elements. The definitive nature ofa complete digital 3D model and the nearly photorealistic rendering capabilities of modern computergraphics systems can mask or hide this degree of uncertainty. It is of vital importance, then, that asystem for sharing scientifically authenticated cultural heritage models directly facilitates accurate,conspicuous documentation of the uncertainties inherent in the models (the recent London Charter[2006] establishes principles for such transparency in 3D visualization of cultural heritage). Examplesof the types of uncertainties in such models include the following.

—Structural Architecture. For example, did a wall of a church have three or four windows?

—Geometric Dimensions. Perhaps the wall of the church is known to be between 8 and 12 meters wide,but no further information is known.

—Stylistic Features. Was the door frame of the church decorated in a Romanesque or Baroque style?

—Temporal Correspondence. When was the east side of the church expanded to accommodate a newchapel?

—Construction Materials. Was the west exterior wall stucco or brick?



The 3D model creation process can yield models that may invalidly depict specific answers to thesetypes of questions, even though there is not enough actual knowledge to support such a reconstruction.

To date, there has been insufficient research confronting the problems of uncertainty visualization in3D virtual environments such as archaeological reconstructions. One research group in Germany in thelate 1990’s anticipated the problem and proposed using nonphotorealistic (sketch-like) rendering stylesfor depicting those portions of 3D models with uncertain features [Strothotte et al. 1999]. Other morerecent work looked at the specific aspect of visualizing temporal uncertainty [Zuk et al. 2005]. Thistopic is ripe for experimentation with a variety of approaches for communicating the different typesof uncertainties that occur in 3D cultural heritage models; a few possible directions for investigationinclude the following.

—A New 3D Symbology (a set of 3D glyphs) for Indicating Uncertainty. Depicting uncertainties in digital3D reconstructions has been difficult in part due to the lack of any such standardized techniques toborrow from, even in traditional nonelectronic and 2D illustration forms. Thus, we should investigatedevelopment of a new set of intuitive 2D and 3D icons that are sufficiently rich to represent the broadrange of the types of uncertainty encountered in 3D cultural heritage models.

—Animation Techniques. Real-time visualization of 3D models allows time-varying depiction techniquesto be used for representing uncertainty. As a simple example, an architectural feature on a modelthat is estimated to be properly located with only a 30% certainty level could be switched on for acorresponding percentage of the time during some fixed interval (its duration chosen so as to minimizeuser distraction).

—Rendering Techniques. Modern computer graphics hardware is powerful enough to render complexstylistic effects, even at full interactive rates. A variety of nonphotorealistic techniques, blurring,transparency, false coloring, and other rendering methods could be experimented with to determineeffective schemes for representing uncertainty of 3D geometric elements.

—Combining Textual Metadata with 3D Visualization. Many hypothetical reconstructions are the resultof highly complex design decisions that may be difficult or impossible to present iconically in the 3Dvisualization, and are best described as traditional textual metadata or 2D images. User interactionparadigms for displaying this textual information appropriately should be explored, that depict themetadata in an uncluttered manner, collocated with the 3D visualization.

4.4 Version Control for 3D Models

Version control (also known as revision control or source control) is an established body of techniquesfor managing multiple revisions of the same unit of information. Version control is most commonlyused in the software development process to manage repositories of source code under development.Revision control systems allow software engineers to track and synchronize modifications, and supportconcurrent work by a large group, as well as development of multiple versions of software derived froma common code base.

The methods of version control as applied to software development can theoretically be adapted toany type of electronic document. There has been some investigation into using version control withCAD systems [Chou and Lim 1986], but no concerted research effort has yet studied the application ofversion control techniques to the development and dissemination of 3D models. Version control shouldbe a fundamental feature of an effective archive for publishing 3D cultural heritage models.

Virtual environments of cultural heritage sites are a particularly demanding class of 3D models formanaging versioning. 3D models of archaeological reconstructions often are changed and evolve overlong periods of time as new knowledge comes to light from excavations and further scholarship. Multiplescholars may hold contrasting theories about the appropriate form of the reconstruction, and generate



multiple versions of the same model that are variations on a common basis. Models themselves maydepict changes across large temporal spans, with multiple versions corresponding to reconstructions atspecific points in time. Additionally, the technological aspects of 3D modeling often require developmentand distribution of multiple versions of models. For example, 3D models constructed for use in fullyinteractive, real-time flythroughs are usually of much lower geometric complexity than models usedfor static, high-quality renderings. Similarly, 3D models may be distributed in a variety of the common3D file formats, to provide for interoperability and usage by a wide spectrum of users. A version controlsystem for 3D models would efficiently manage the creation, storage, and dissemination of all thesepotential versions of cultural heritage models.

Future research, then, should investigate the applicability of traditional revision control methods,and further extend them to be suitable for managing large archives of 3D cultural heritage models. Weanticipate developing solutions for a 3D model version control system with the following features.

—Tracking of Every Addition, Deletion, and Modification to 3D Models. As with source code revisionsystems, each edit should be stored with metadata representing the time of the edit, the identity ofthe editor, and comments or annotations describing the nature of the modification.

—3D Difference Computation and Visualization. Traditional revision control systems for textual docu-ments provide functionality for convenient viewing of the differences between two document versions,similar to the “diff” command available on UNIX systems. We call for investigation of methods fora “3D diff” function that allows rapid 3D visualization of the difference between two versions of a3D model. This capability could use geometric analysis algorithms and information visualizationtechniques to automatically identify and then visually highlight the variations among 3D models.Similar techniques for 3D model comparison have been applied before in 3D digital heritage analy-sis [Niederost 2001].

—Delta Compression for 3D Models. For purposes of storage efficiency, revision control systems usuallyretain only the differences between successive or branched versions of documents; similar techniquesfor delta compression of 3D geometric components should be investigated.

4.5 Provision for Application of Analytical Tools to 3D Models

Cultural heritage 3D models are most commonly regarded as means for documenting and visualizingof historic sites and artifacts. While these purposes themselves are valuable objectives, we should hopeto go beyond visualization and further extend the value of the archived 3D data by facilitating theapplication of powerful digital analytical tools to the 3D models disseminated via our archives.

Examples of applying analysis tools to 3D cultural heritage models include algorithms for automatedreassembly of digitized fragmented objects [Koller and Levoy 2006], or GIS analysis of the spatialdistribution of excavated artifacts. Analysis tools for 3D models of reconstructed architectural sitesmight measure and analyze, for example, the functionality of buildings in terms of room capacity,circulation, illumination, and ventilation. Simulation software for computing illumination and air flowin building interiors already exists, so research efforts should concentrate primarily on developinginformation architectures and user interfaces such that analytical tools can be efficiently applied in ageneral way across a wide range of the 3D models to be archived in virtual heritage repositories.

4.6 Indexing and Searching 3D Archives

Other critical functions associated with digital archives of 3D cultural heritage models that requiresfurther research are indexing and searching. As an increasing number of virtual heritage sites are dig-itized and modeled in 3D, archives will grow to the point where powerful retrieval tools are necessaryfor users to quickly find the information they seek. An active area of research that is directly relevant is



shape-based retrieval methods, that measure shape similarity between 3D models [Funkhouser et al.2003; Tangelder and Veltkamp 2004]. We expect that the rich textual annotation required of scien-tifically authenticated 3D cultural heritage models will often provide appropriate metadata for textmatching and effective retrieval, but research results have demonstrated the value of using multiclas-sifiers for large 3D databases that match on both text and 3D shape simultaneously [Min et al. 2004].Thus we encourage further research into algorithms for indexing and searching on inherent modelattributes that do not depend on human annotation. The ability to search model archives based on 3Dshape queries can also be useful in many cultural heritage analysis processes that use 3D models, suchas reassembly of fragmented artifacts.

4.7 Interoperability

The creation of 3D cultural heritage archives also requires new solutions for interoperability of 3Dmodels, both with other models in the archive, as well as with external sources. For example, effectivearchives should support conversion of models to a common format, and provide access to georeferenc-ing metadata that allows different models to be properly located relative to one another in the samecoordinate system. Users need to be able to find, retrieve, and combine virtual environment modelsfrom different sources. A model of the Roman Forum in 400 A.D. made by one team with MultiGenCreator should ideally interoperate with a model of the neighboring Forum of Julius Caesar madeby another team using 3D Studio Max as their modeling software. Models from 3D virtual heritagerepositories should also be compatible with centralized common 3D geospatial visualization systemssuch as Google Earth, which are quickly gaining popularity as platforms for presenting geolocatedinformation.

These needs are, of course, by no means limited to 3D model archives. They loom large throughoutdigital culture, as has been recognized by the European Union in an influential study by Mulrenin andSzauer [2002, page 186]:

From a technological point of view, providing seamless, integrated access to services and cul-tural heritage resources is primarily a question of convergence and interoperability. At thebasic level, convergence and interoperability mean providing users with the capacity to treatmultiple digital collections in various cultural institutions as one. For cultural heritage insti-tutions, one of the primary obstacles to providing integrated seamless access over computernetworks relates to the fact that over the Internet, protected databases or online catalog[s]cannot be directly searched.

4.8 Long-Term Preservation of 3D Data

One of the most pressing needs for 3D cultural heritage archives is ensuring the survivability of themodels [Frischer et al. 2002]. More and more 3D models are being created by digital archaeologists andarchitectural historians around the world, but no one is collecting them, let alone addressing ways ofpreserving them. In a sense, the activity of 3D modeling can be viewed as an integral part of conservationscience, and, indeed, in a recently proposed ICOMOS charter on archaeological site presentation, threearticles explicitly discuss the appropriate uses of computer models as substitutes for restoration andanastylosis [Frischer and Stinson 2003]. Anecdotal evidence suggests that even digital models that areonly two or three years old are losing their original functionality and information richness because ofpoor archival practice. CAD software products used to create virtual heritage models belong to a rapidlychanging market segment, which means that updated versions are inevitable (on the order of every 18months) and full reuse of data (without loss of information) is not secured at all.



Once again, the problem transcends 3D models. As Mulrenin and Szauer note [2002, page. 210]:

Not brittle papyrus and crumbling mortar is the most severe threat to our cultural heritagetoday, but, as Mary Feeny expressed it, “the death of the digit” [Feeny 1999]. This ‘death ofthe digit’ is related primarily to two factors that put at jeopardy current efforts in archivingand preserving our digital cultural heritage: first, technology develops ever more rapidly,reducing the time before a particular technology becomes obsolete; and secondly, unlike theiranalogue counterparts, digital resources are much more ‘unstable’ with the effect that theintegrity and authenticity of digital cultural resources is corrupted.

Hence Mulrenin and Szauer [2002] recommend to the European Union that “the worst thing to do isadopting a ‘sit and wait’ attitude. Instead, cultural heritage institutions should actively approach thepreservation issue and try to take as much control as possible. This means to actively manage the lifecycle of digital resources, from data creation and data management to data use and rights management.”We agree with the general policy propounded by Mulrenin and Szauer [2002] and propose the applicationof it to the specific case of proactive approaches to the collection, categorization, dissemination, andpreservation of 3D models of cultural heritage sites.

The digital libraries community has made strong progress in technical solutions for preservation andreliability of access, and we recommend extending and applying these techniques to 3D archives. TheLOCKSS system [Rosenthal and Reich 2000], for example, has proven highly effective for preservationof Web-published academic journals, and it would be fruitful to consider the ramifications of a similarpeer-to-peer, highly replicated architecture for 3D archiving, along with appropriate implementationsof transparent format migration or emulation schemes.

4.9 Organizational and Logistical Structure of a Peer-Reviewed 3D Archive

The existing model of traditional peer-reviewed print journals provides the inspiration for scientificallyauthenticated online 3D cultural heritage archives. Further research is necessary, however, to study anddevelop the most appropriate organizational, logistical, and sociological structures for curation of thisnew type of venue for dissemination of research results. As a first step, we anticipate that the traditionalstructure with an appointed editorial board is suitable, with editors forming committees of specializedreviewers to peer review submissions of 3D models to the archive. The editorial board would consist ofexpert scholars in the relevant subject domains (e.g., classical architecture, art history, 3D digitization).A standardized report form would be designed for the archive, that would be published with each 3Dmodel. Other logistical issues to be explored would include licensing and copyright arrangements for3D models, and schemes by which the 3D model developers could receive revenue from users who accessmodels from the archive, when desired.

5. CONCLUSION

Successfully addressing these research challenges will dramatically improve the state of the art in themanagement and utility of 3D cultural heritage model archives. Scholarship in nearly all fields of thehumanities and sciences will be empowered by having reliable, sustained access to authenticated 3Dmodels. The realization of a peer-reviewed resource for publishing 3D models and clearly recognizingtheir validity will further encourage the rapid growth and acceptance of virtual heritage as a compellingnew manner of communicating knowledge.

Resolving these issues will involve an unusually broad multidisciplinary research effort. The technicalcomputer science aspects themselves span the fields of computer security, computer graphics, databases,



computational geometry, visualization, and information retrieval, to name only a fraction. Domainexpert humanist scholars will be necessary to define the scientific and technological user needs of thevirtual heritage community, such as specifying metadata architectures appropriate for their objects ofstudy.

At the University of Virginia, our long-term objective is to create a 3D cultural heritage archive thatfulfills the vision outlined here. Our multi-year SAVE initiative (“Serving and Archiving Virtual Envi-ronments”) has been supported by the National Science Foundation to tackle the associated researchproblems, with an initial focus on development of digital rights management methods for protectingand sharing 3D models, as well as conducting a needs assessment of potential contributors to an open3D archive. Currently we are studying the problems of uncertainty visualization and 3D versioning,and we anticipate the implementation and deployment of a pilot version of a peer-reviewed SAVE 3Dcultural heritage archive that will give us a testbed to begin addressing the more systems-orientedand organizational research challenges. We look forward to tackling these problems, along with otherresearchers in the computing and cultural heritage community!

REFERENCES

ARNOLD, C., HUGGETT, J., REILLY, P., AND SPRINGHAM, C. 1989. Mathrafal: A case study in the application of computer graphics. In

Proceedings of the Conference on Computer Applications in Archaeology (CAA), S. Rahtz and J. Richards, Eds. Vol. 16. 147–156.

BARCELO, J., FORTE, M., AND SANDERS, D., EDS. 2000. Virtual Reality in Archaeology. ArcheoPress, Oxford, UK (British Archae-

ological Reports, International Series #843).

BIEK, L. 1985. LERNIE XIV: Comparology and stereovideo. In Proceedings of the Conference on Computer Applications inArchaeology (CAA). Vol. 12. 1–35.

BUCKLAND, P. AND WILCOCK, J. 1973. An experiment in the use of a computer for on-site recording of finds, and the use of remote

terminals for archaeological site records. In Proceedings of Computer Applications in Archaeology 1, Science and Archaeology(CAA73). Vol. 9. George Street Press, Stafford, UK.

CHOU, H. AND LIM, W. 1986. A unifying framework for version control in a CAD environment. In Proceedings of the 12thInternational International Conference on Very Large Data Bases (VLDB’86). 336–344.

CLARK, J. T., BERGSTROM, A., LANDRUM, J., LARSON, F., AND SLATOR, B. 2002. Digital archive network for anthropology (DANA):

Three-dimensional modeling and database development for Internet access. In Proceedings of the 12th International Sympo-sium on Virtual Reality Archeology and Cultural Heritage (VAST’2000). British Archaeological Reports, F. Niccolucci, Ed. Vol.

834.

CYARK. http://www.cyark.org.

DIGITAL HAMMURABI PROJECT. http://www.jhu.edu/digitalhammurabi/.

FEENY, M., ED. 1999. Digital Culture: Maximising the Nation’s Investment. National Preservation Office, London, UK.

FORTE, M. AND SILIOTTI, A. 1997. Virtual Archaeology. Abrams, New York.

FRISCHER, B. 2003. Mission and recent projects of the UCLA cultural virtual reality laboratory. In Proceedings of the ConferenceVirtual Retrospect. R. Vergnieux and C. Delevoie, Eds. 65–76.

FRISCHER, B. 2008. From digital illustration to digital heuristics. In Beyond Illustration: 2D and 3D Digital Technologies asTools for Discovery in Archaeology, B. Frischer and A. Dakouri-Hild, Eds. Vol. 1805. BAR International Series, v–xxiv.

FRISCHER, B., ABERNATHY, D., GIULIANI, F. C., SCOTT, R., AND ZIEMSSEN, H. 2006. A new digital model of the roman forum. J.Roman Arch. Suppl. Series 61.

FRISCHER, B., NICCOLUCCI, F., RYAN, N., AND BARCELO, J. 2002. From CVR to CVRO: The past, present, and future of cultural

virtual reality. In Proceedings of the International Symposium on Virtual Reality Archeology and Cultural Heritage (VAST’00).British Archaeological Reports, F. Niccolucci, Ed. Vol. 834. 7–18.

FRISCHER, B. AND STINSON, P. 2003. The importance of scientific authentication and a formal visual language in virtual models

of archaeological sites: The case of the House of Augustus and the Villa of the Mysteries. In Proceedings of the Ghent Conferenceon Heritage, Technology, and Local Development. D. Callebaut, Ed.

FUNKHOUSER, T., MIN, P., KAZHDAN, M., CHEN, J., HALDERMAN, A., DOBKIN, D., AND JACOBS, D. 2003. A search engine for 3D models.

ACM Trans. Graph. 22, 1, 83–105.

JOHNSON, C. AND SANDERSON, A. 2003. A next step: Visualizing errors and uncertainty. IEEE Comput. Graph. Appl. 23, 5, 6–10.



KOLLER, D. AND LEVOY, M. 2006. Computer-Aided reconstruction and new matches in the forma urbis romae. Bull. DellaCommissione Archeol. Comunale di Roma.

KOLLER, D., TRIMBLE, J., NAJBJERG, T., GELFAND, N., AND LEVOY, M. 2006. Fragments of the city: Stanford’s Digital Forma Urbis

Romae Project. J. Roman Archaeol. Suppl. 61, 237–252.

KOLLER, D., TURITZIN, M., LEVOY, M., TARINI, M., CROCCIA, G., CIGNONI, P., AND SCOPIGNO, R. 2004. Protected interactive 3D graphics

via remote rendering. ACM Trans. Graph. 23, 3, 695–703.

LEVOY, M., PULLI, K., CURLESS, B., RUSINKIEWICZ, S., KOLLER, D., PEREIRA, L., GINZTON, M., ANDERSON, S., DAVIS, J., GINSBERG, J., SHADE,

J., AND FULK, D. 2000. The digital Michelangelo project. In Proceedings of ACM SIGGRAPH International Conference onComputer Graphics and Interactive Techniques. 131–144.

LONDON CHARTER. 2006. The London Charter for the use of 3-dimensional visualisation in the research and communication of

cultural heritage. www.londoncharter.org.

MACEACHREN, A. M., ROBINSON, A., HOPPER, S., GARDNER, S., MURRAY, R., GAHEGAN, M., AND HETZLER, E. 2005. Visualizing geospatial

uncertainty: What we know and what we need to know. Cart. Geogr. Infor. Sci. 32, 3, 139–160.

MIN, P., KAZHDAN, M., AND FUNKHOUSER, T. 2004. A comparison of text and shape matching for retrieval of online 3D models. In

Proceedings of the European Conference on Digital Libraries. 209–220.

MULRENIN, A. AND SZAUER, A. 2002. The DigiCULT Report. Technological Landscapes for Tomorrow’s Cultural Economy. Un-locking the Value of Cultural Heritage. The European Commission, Directorate-General for the Information Society.

NIEDEROST, J. 2001. 3D reconstruction and accuracy analysis of historical relief models. In Proceedings of the 3rd InternationalSeminar on New Development in Digital Photogrammetry. 34–37.

OHBUCHI, R., MUKAIYAMA, A., AND TAKAHASHI, S. 2002. A frequency-domain approach to watermarking 3D shapes. Comput.Graph. Forum 21, 3.

PANG, A. 2001. Visualizing uncertainty in geospatial data. In Proceedings of the Workshop on the Intersections BetweenGeospatial Information and Information Technology.

PARIS, C. AND FIEDLER, M., EDS. 2005. Protection of stored data by system level encryption schemes. EuroNGI Deliverable

D.WP.JRA.6.3.5.

PRAUN, E., HOPPE, H., AND FINKELSTEIN, A. 1999. Robust mesh watermarking. In Proceedings of ACM SIGGRAPH InternationalConference on Computer Graphics and Interactive Techniques. 49–56.

ROSENTHAL, D. S. AND REICH, V. 2000. Permanent Web publishing. In Proceedings of the USENIX Annual Technical Conference,Freenix Track. 129–140.

SANDERS, D. 2006. Personal communication.

SHIAW, H., JACOB, R. J. K., AND CRANE, G. R. 2004. The 3D vase museum: A new approach to context in a digital library. In

Proceedings of the 4th ACM/IEEE-CS Joint Conference on Digital Libraries. 125–134.

STROTHOTTE, T., MASUCH, M., AND ISENBERG, T. 1999. Visualizing knowledge about virtual reconstructions of ancient architecture.

In Proceedings of the Computer Graphics International Conferences. 34–43.

TANGELDER, J. AND VELTKAMP, R. 2004. A survey of content based 3D shape retrieval methods. In Proceedings of Shape ModelingApplications Conference. 145–156.

TURBOSQUID. http://www.turbosquid.com.

VERGAUWEN, M. AND GOOL, L. V. 2006. Web-Based 3D reconstruction service. Mach. Vis. Appl..VERYKIOS, V., BERTINO, E., FOVINO, I., PROVENZA, L., SAYGIN, Y., AND THEODORIDIS, Y. 2004. State-of-the-Art in privacy preserving

data mining. ACM SIGMOD Rec. 3, 1, 50–57.

ZHANG, L., TONG, R., SU, F., AND DONG, J. 2002. A mesh watermarking approach for appearance attributes. In Proceedings ofPacific Graphics.

ZUK, T., CARPENDALE, S., AND GLANZMAN, W. 2005. Visualizing temporal uncertainty in 3D virtual reconstructions. In Proceedingsof the 6th International Symposium on Virtual Reality, Archaeology and Cultural Heritage (VAST’05). 99–106.

Received January 2007; accepted July 2008


Date post:	04-Dec-2020
Category:	Documents
Upload:	others
View:	2 times
Download:	0 times

Artifacts and Original Intent: A Cross-Cultural...

Documents