Bruce Edmonds • Ruth Meyer
Editors
Simulating Social Complexity
A Handbook
�Springer
Editors
Bruce Edmonds Ruth Meyer CPM- Centre for Policy Modelling Manchester Metropolitan University Manchester United Kingdom
ISSN 1860-0832
ISB 978-3-540-93812-5
DOl I 0.1007/978-3-540-93813-2
ISSN 1860-0840 (electronic) ISBN 978-3-540-93813-2 (eBook)
Springer Heidelberg New York Dordrecht London
Library of Congress Control Number: 2013934710
c Springer-Verlag Berlin 1-leidelberg 20 I :l
Thi> work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned. specifically the rights of translation. reprinting, reuse of illustrations. recitation. broadcasting. reproduction on microfilms or in any other physical way. and transmission or information storage and retrieval, electronic adaptation. computer software. or by similar or dissimilar
methodology now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection with reviews or scholarly analysis or material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work. Duplication of this publication or parts thereof is permitted only under the provisions of the Copyright Law of the
Publisher"s location. in its cun·ent version, and pennission for use must always be obtained from Springer. Permissions for use may be obtained through RightsLink at the Copyright Clearance Center. Violations are liable to prosecution under the respective Copyright Law. The use of general descriptive names, registered names. trademarks. service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. While the advice and information in this book are believed to be true and accurate at the date of publication. neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made. The publisher makes no warranty. express or implied, with respect to the material contained herein.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)
Chapter 10
Participatory Approaches
Olivier Barreteau, Pieter Bots, Katherine Daniell, Michel Etienne,
Pascal Perez, Cecile Barnaud, Didier Bazile, Nicolas Becu,
Jean-Christophe Castella, William's Dare, and Guy Trebuil
0. Barreteau (G) IRSTEA. UMR G-EAU, 36 1 rue Jean-Fran«ois Bre!On, B P 5095. Montpellicr 34 1 96. France e-mai I : ol i vier.barreteau@i rstea. fr
P. Bots Delft University of Technology, Faculty of Technology. Policy and Managemenl. Delfl 50 1 52600 GA, The Netherlands e-mai l : p.w.g.bots@tudel ft .nl
K . Daniell Centre for Policy I nnovation. The Australian National University, Canberra ACT 0200, Australia e-mai l : katherine.daniel [email protected]
M. Etienne Ecodevelopment Uni tDomaine St-Paul, site Agroparc, Avignon cedex 9 849 1 4, France e-mail: et ienne@avignon. inra.fr
P. Perez SMART, University of Wollongong, Northfields Ave, Wol longong NSW 2522, Australia e-mai l : [email protected]
C. Barnaud I N RA, Centre I NRA de Toulouse U M R Dynafor Chemin de Borde Rouge. Castanet Tolosan Cedex B P 526273 1 326, France e-mai I: cecile.barnaud@toulouse. inra.fr
D. Bazile • W. Dare • G. Trebuil Cirad GREEN, TA C-47/F. Campus international de Baillarguet, Montpe l l ier Ccdex 5 34398, France e-mai l : [email protected]: [email protected]: [email protected]
N. Becu CNRS. Laboratoire de geographic PRODIG 2. rue Vale!te, Paris 75005. France e-mai l : nicolas.becu@univ-paris l . fr
J . -C. Castella l nst itut de Recherche pour le Developpement ( I RD). U M R 220 G R ED ( I RD UPV Montpe l l ier 3 ). PO Box 5992, Vient iane, Laos e-mai l : [email protected]
B. Edmonds and R. Meyer (eds . ) . Simularing Social Comple.rirv, Understanding Complex Systems. DOl I 0. 1 007 /978-3-540-938 1 3-2_ 1 0. 1(" Springer-Verlag Berl in Heidelberg 20 1 3
1 97
1 98 0. Barreteau et al .
Why Read This Chapter? To help you understand how one might i nvolve stakeholders in all stages of the model l ing process. This approach al lows for
including stakeholders' expertise as wel l as giving them more control over the process.
Abstract This chapter aims to describe the diversity of part icipatory approaches in
relation to social s imulations, with a focus on the interact ions between the tools and
part ic ipants. We consider potential interactions at all stages of the model l ing
process: conceptual design; implementat ion; use; and s imulat ion outcome analysis .
After reviewing and class ifying ex ist ing approaches and techniques, we describe
two case studies with a focus on the integration of various techniques. The first case
study deals with fire hazard prevention in southern France, and the second one with groundwater management on the Atol l of K iribati . The chapter concludes with a
discussion of the advantages and l im itations of part icipatory approaches.
10.1 Introduction
In this chapter, social s imu lation is cross-examined with a currently very active
trend in pol icy making: part ic ipation or stakeholder involvement. This cross
examination has two main outputs: the development of tools and methods to
improve or faci l i tate pat1 icipation; and the deve lopment of more grounded simula
t ion models through part ic ipatory model l ing. Technological development provides
new devices to fac i l itate interaction around s imulation mode ls : from the phase of
conceptual design to that of pract ical use. In many f ields there is a growing
requirement from stakeholders and the public to become more act ively involved
in pol icy making and to be aware of probable changing trends due to global pol icy
dec isions. New tools and methods related to social s imu lation have started to be
made avai lable for this purpose such as many Group Dec ision Support Systems
which use computer s imulation, inc luding potential l y social i tems components, to
faci l itate communication to formulate and solve problems col lectively ( DeSancti s and Gal lupe 1 987; Shakun 1 996; Whitworth et a l . 2000) . I n addi t ion, s imulat ion of
soc ial complexity occurs in models whose val idation and suitabi l i ty depend on their close fit to society, as well as on their acceptab i l i ty by it . These i ssues are tackled
through the use of participatory mode l l ing, such as group model bui lding ( Vennix
1 996) or part ic ipatory agent based s imulat ions ( Bousquet et al . 1 999; Guyot and
Honiden 2006; Moss et a l . 2000; Pah l-Wostl and Hare 2004; Ramanath and G ilbert
2004). The topic is also related to part ic ipatory design as it is a mean of involving
end-users of computer systems in their design, including social s imulations
focussed ones (Schuler and Namioka 1993) .
Group Decision Support as wel l as Participatory Model l ing stem from the
interactions between simulation models and participants. There is a d iversity of
ways though which these interactions m ight take place. They are related to the
diversity of approaches to simu late society or to organise participation. It is important to make the choices made for these interactions explicit : for dist inction between
approaches to be possible; to provide the opportunity for stakeholders to discuss the
I 0 Part ic ipatory Approaches 1 99
process; and for them to be prepared to be involved in . There is a need to go further
than the development of tools as they are l iable to create f i lters that reshape the understanding of soc ial complexity. Description of the mechanisms behind
interactions is a way to qual i fy the potential effects of these interact ions. This chapter aims to describe the diversity of patticipatory approaches in
relat ion to social s imu lat ions, with a focus on the interactions between the tools
and part icipants. This overview is l imi ted to s imulat ion models. Model is considered here as a representat ion of shared knowledge, which means the gathering of
pieces of knowledge and assumpt ions about a system, written altogether in a model
so that they m ight play or work together. We l imi t th is scope further to simulation
mode l , hence models inc luding the representation of dynamics. We consider here
potential interact ions among part icipatory and mode l l ing processes at al l stages of the model l ing process: conceptual design; implementat ion; use; and simulation
outcome analysis .
The f i rst section of this chapter out l ines a number of factors which have paved
the way for development of the association between soc ial simu lation and part ic i
pat ion. There is a large body of l i terature in which authors have developed their
own participatory model l ing approaches, j ustified by some speci f ic expectat ions on
part ic ipation for model l ing or vice-versa. This first sect ion makes a synthesis of
these expectat ions and draws out some principles on which various part icipatory
mode l l ing sett ings should be assessed. The second section describes some existing
techniques and approaches. The third section proposes a class i ficat ion of these
part icipatory approaches according to three d imens ions: the level of involvement in the process; the t imel iness of involvement; and the heterogeneity of population
i nvolved. The fourth section describes two case studies with a focus on the integration of various techniques. We discuss the advantages of these approaches
but also some l imi ts, according to the expectations and in comparison wi th more
tradit ional techniques in the f i fth section.
10.2 Expectations of Using Participatory Approaches
with Simulation of Social Complexity
Joint use of part icipatory approaches with social s imulat ions is based upon three
categories of expectations. They vary according to the target of the expected
benefits of the assoc iation:
I. Qual i ty of the simu lation model per se;
2 . Suitabi l i ty of the simulation model for a given use; and
3. Patt icipation support.
These three targets are l inked to three di fferent components of a model l ing
process. Target one is l inked to the output, target three to the source system, and
target two to the relation between both the output and source system. I n this section
we further develop these three categories.
200 0. Barreteau et al .
10.2.1 Increasing Quality of Simulation Models of Social
Complexity
The objective here i s to produce a good qual i ty model to simulate social complexity.
Partic ipation is then pragmat ically assumed to be a means for improv ing this
qual i ty . There is no normative be l ief which would value part icipation by itse l f in
this category of expectations.
Qual i ty of the simulat ion model i s understood here rather classical ly with the
fol lowing indicators:
- Real ism: is the s imulat ion model able to tackle key features of the social
complexity i t aims to represent?
- Efficiency: is the simulation model representing its target system with a mini
mum of assumptions and minimal simulat ion run-times?
Qual i ty of the representation according to its use is another c lassical indicator of
a simulat ion model's qual i ty. lt is specifical ly tackled in the following subsection.
10.2.1.1 Taking Social Diversity and Capacity to Evolve into Account
One of the key features to be taken into account when representing a social system is
to deal wi th i ts diversity. This d iversity i s related not only to individual
characteri st ics, but also to v iewpoints, expectations towards the system, and positions
in the decision making processes. Deal ing with diversity in simulation of social
complexity involves embracing it as well as profiting by i ts existence.
Classically, deal ing with diversi ty is a process of aggregation or select ion.
Aggregation consists of the ident i fication of classes of indiv iduals and
representati ves for them. Selection consists of choosing a few cases with a l l of
their characteristics . This may lead to very simple simulation models with a generic
diversity. Aggregation is rather greedy on data and model l i ng t ime and is still
dependent on the v iewpoint of the observers who provide the information leading to
the categorisat ion. Selection is weak to cope with relations among various sources
of d iversity.
I nvolvement of stakeholders in the model l i ng process al lows them to bring their
own diversity. Concerns over representation are then transferred onto the consti tu
t ion of the sample of part icipants. Fischer and colleagues have shown through
development of si tuations to support creativity in various fields, such as art, open
source development and urban planning, that diversity, as well as complexity, i s
important to enhance creat iv i ty ( Fi scher et al . 2005) . This creativ i ty i s expected to
pave the way for surprises in the simu lation model .
Involvement of stakeholders in the model l ing process is a way to external ise part
of this diversity outside the model towards a group of stakeholders. The i ssue is then
to work on the relation between the model and a number of stakeholders to allow a
transfer of knowledge and ideas.
1 0 Part ic ipatory Approaches 20 1
Social systems are open and evolving. Their definition depends on the viewpoint
of the analyst. As far as simulation i s concerned, this means depending on the
viewpoi nt of the model designer(s ) . This choice means framing: cutting a number
of l inks around the boundaries of the system studied, as well as around the
interpretation which might occur based on the s imulat ion outcomes ( Dewulf et al.
2006). First ly, part ic ipation provides the opportunity to consider problem
boundaries which would be plural ly defined, increasing the potential coherence of
the model . However, i t is st i l l an operation of cutt i ng l inks out of the real world
situation, even though these chosen cutt i ngs are more grounded and discussed.
Secondly, interactive use of a simu lation model i s a means to keep some of these
l inks open and active, with part icipants as driving belts . Stakeholders are embedded
in social networks which cross the boundaries into the physical and environmental
networks. They make the l inks come al ive, which al lows them to function and be
updated.
There is thus a need to question the boundaries set in the interactive setting:
actors in the neighbourhood; concerns of actors connected to those tackled by the
(s imulation) mode l ; and how these relations are to be mobi l i sed in the interaction.
10.2.1.2 Distribution of Control
A key characteristic of social systems which is to be addressed through social
s imulat ion is their complexity. This complexity leads to various consequences, such
as the emergence of phenomena, delay effects or discont inuities in some trends,
which are present in social systems as in any complex systems. These are usual ly
the effects which one l ikes to discover or better understand when experimenting
with social simulations. From the internal point of view of simu lations, Sche l l ing
has shown experimental ly that reproducing sett i ngs with mult iple dec ision centres
improves the qual i ty of representation of complexi ty (Schel l ing 1 96 1 ). He could generate complexity through experimental games because of the presence of
independent decision centres, the players. This result has also been shown with
simulat ions used for forecasting (Green 2002 ) . Green compared the capacity of
forecasting the outcome of past social confl icts with: a role playing game with
students; game theorists; and a group of experts . He compared the simulated outcomes with those from the real negot iations and found that the role playing
game sett ing produced the best results . This was the one with the main distribution
of dec isions among autonomous centres.
The purpose of associating part icipatory processes and social s imu lation here is
then to increase the complex ity through interactive use or implementat ion of a
social mode l . Unless computational agents are effect ively used, which is rare
( Drogoul et al . 2003) , formal theories of complex systems that are completely
embedded in a simulat ion model do not s imulate complex patterns but implement
an explanat ion of a complex pattern. In other words, they should be implemented in
a distributed setting with autonomous entit ies. Part icipatory approaches provide
202 0. Barreteau et al .
such settings. There is then an issue of a deep connection between a simulation
model and partic ipants in a part ic ipatory mode l l i ng sett ing.
10.2.2 Improving Suitability of Simulation Model's Use
Quality of a model is a lso assessed according to i ts su itabi l ity for its intended use. I n
t h i s subsection, two cases of use are considered: knowledge increase; and policy
making. In both cases, i t i s expected that involvement of stakeholders at any stage
of a model l ing process wi l l aid better tun ing of the model with its i ntended use :
either through interactions with people represented in the model , or with potential
users. Both cases have a major concern with making viewpoints explicit .
10.2.2.1 Case of Increasing Knowledge
The case of use for knowledge increase bui lds upon the previous subsection. The
key element treated here deals with the uncertainty of social systems. The involve
ment of stakeholders represented in the simulat ion model is a way to improve i ts
val idation or calibrat ion. Partic ipants may bring their knowledge to reduce or better
qual i fy some uncertainties. The simulation model is then expected to give back to the part ic ipants simulation outputs based on the interactions between their pieces of
knowledge. On the other hand, this feedback is sometimes di fficult to val idate
( Manson 2002 ) . I ts presentation and discussion with stakeholders represented in the
simulation model is a way to cope with this issue. This approach has been explored by Barreteau and col leagues to improve the val idation of an Agent Based Model of
i tTigated systems in Senegal River val ley ( BatTeteau and Bousquet 1 999) . The
format of th is feedback, information provided and medium of communication,
might make the model real ly open to discussion. Thi s joins another expectation which i s probably the most common in work that
has so far implemented such part icipatory approaches with a social s imulat ion model : making each part ic ipant ' s assumptions expl ici t , included the modellers
( Fischer et al. 2005 ; Moss et al. 2000; Pahl-Wostl and Hare 2004) . This is a
requirement from the s imulation mode l l i ng community: mak ing stakeholders'
bel iefs, points of view and tac i t knowledge explici t ( BatTeteau et al. 200 1 ; Cockes
and lve 1 996; D' Aquino et al. 2003; McKinnon 2005 ) . Moreover, so that
part ic ipants m ight become part of the model , the assumptions behind the model
should be made expl icit in order to be discussed, as should the outputs of the
simulat ions so that they can also be discussed, transferred and translated in new
knowledge. This is to overcome one major pitfal l identified with the development of
mode ls which i s the under-use of decision support models because of their opacity ( Loucks et al . 1 985; Reitsma et al . 1 996) . Thi s concern of making explici t
assumptions in the model l i ng process is also at the heart of the part ic ipatory
approach community. One aim of gathering people together and making them
I 0 Part icipatory Approaches 203
col lectively discuss their si tuation in a pat1icipatory sett ing is to make them aware
of others' viewpoints and interests. This process involves and st imulates some
explanation of tacit posit ions.
This means that the interactive sett ing shou ld allow a bi-directional transfer of
knowledge between stakeholders and the simulation mode l : knowledge el icitation
in one direction and validation and explanation of s imulat ion outputs in the other
direct ion.
10.2.2.2 Case of Policy Making
ln the case of simu lation focusing on pol icy issues, there is a pragmatic, moral, and
now sometimes legal need to involve stakeholders, which may lead to open the
black box of models of social complexity used in policy making. Post-normal approaches aim at making the dec ision process and its tools expl icit so that
stakeholders can better discuss i t and appropriate its outcomes. When this decision
process involves the use of dec ision support tools, which might inc lude social
s imulat ion models, th is means that the models themselves should be opened to
stakeholders ( Funtowicz et al . 1 999). A simulat ion model is then expected to be
explicit enough so that stakeholders who m ight be concerned by the implementat ion of the pol icy at stake could discuss it. This legi t imisation is social ly based,
while val idation, as ment ioned with the previous case of use, is scient i fical ly based
( Landry et a l . 1 996) . Even though val idat ion is sti l l required in this case of use,
because i t is the mode of evaluat ion for some participants, it is rather the
legit imisation of the model by the stakeholders which is to be worked out.
Part icipatory approaches may be a means for opening these models to
stakeholders, provided that fom1ats of communication of models' assumptions and
structure can be genuinely discussed. involvement of stakeholders is expected to raise
their awareness of the assumptions of the model and potentially able to discuss these
and modify them. This includes the evolution of underlying values and choices made
in the design of model.
10.2.3 Simulation as a Means to Support Participation
Social simulat ion might also benefit to pat1icipat ion. Whi le the previous subsection
was dedicated to appropriateness between the model and its use as a group decision
support tool, we focus here on part ic ipation which might be a component of a
dec ision making process.
Social s imulation i s seen here as an opportunity to foster part ic ipation and cope
with some of its pitfal ls ( Eversole 2003) . Use of simu lation mode ls may lead to
some outcomes such as community building or soc ial learning.
204 0. Barreteau et al.
10.2.3.1 Dynamics and Uncertainties
Social systems have to deal with uncertainties just as soc ial simulat ion models do.
This m ight hamper part icipatory processes: in w icked problems (R ittel and Webber
1 973 ) , encountered in many situations where part ic ipatory processes are organised,
stakeholders always maintain the opportunity related to these uncertainties to
challenge others' viewpoints or observat ions. As an example: origin, flow and
consequences of non point source pol l ut ion are uncertain. This leads some farmers
to challenge the accusation, made by domestic water companies downstream of
their fields, that they are pol lut ing their sources. Sometimes, disparate viewpoints
do not conflict. The gathering of these disparate pieces of knowledge i s a way to
reduce uncertainty and al low the group of stakeholders involved in a participatory
process to progress; provided that they can work together.
Another charactetist ic of any social system which m ight hamper part ic ipation is
its dynamicity. Socio-ecological systems exhibit a range of dynamics; not only social , but also natural, which evolve at various paces. ln the appl icat ion developed
by Etienne and colleagues in Causse Mejan, pine tree d i ffusion has a typical t ime
step of 20 years which is long according to the typical t ime steps of land use choices
and assessment (Etienne et al. 2003 ) . In a part ic ipatory process it might be d ifficult to put these dynamics on the agenda. S imulation models are known to be good tools
to deal with dynamic systems.
Simulat ion models are therefore a means to gather distributed pieces of knowledge among stakeholders and to cope with scenarios in the face of uncettainties. They can also help make the part ic ipants aware of potential changes or regime
shifts generated by their interact ions ( Kinzig et al. 2006) .
10.2.3.2 Towards Social Learning
Participation is often l inked with the concept of social learning ( Webler et al . 1 995) .
However, for soci al learning to occur, partic ipants should have a good understand
ing of their interdependencies as well as of the system 's complexity. Social
simu lation can provide these bases, provided that the communication is wel l
developed ( Pahl-Wostl and Hare 2004) .
Thi s learni ng comes from exchanges among stakeholders involved in the par
t ic ipatory process but al so from new knowledge which emerges in the interaction.
External i sation of taci t knowledge in boundary objects (Star and Griesemer 1 989)
is useful for both: i t fac i l itates communication in giving a joint framework to make
one's knowledge expl ic i t ; and it enhances individual, as well as social , creativity
( Fischer et al . 2005) .
S imulat ion mode ls are good candidates t o become such boundary objects. Agent
based models have long been considered as blackboards upon which various
discipl ines could cooperate ( Hochman et al. 1 995) . Through simulation outputs,
they provide the necessary feedback for reflexivity, be it individual or col lective.
1 0 Part icipatory Approaches 205
The question then remains whether such models constrain the format of knowl
edge which m ight be external ised.
10.2.4 Synthesis: A Key Role of the Interaction Pattern
Between Model and Stakeholders
These three categories of expectat ions have led to speci fic requests for the devel
opment of part ic ipation in relation to social simu lation mode ls . In the fol lowing
section, we provide an overview of these techniques. On the basis of the previous
requests, these techniques and methods have to be analysed according to the
fol lowing dimensions:
Set of connections between the part ic ipation arena and simu lat ion model : its
structure, i ts content, and organisation of i ts mobi l i sation;
Control of the process; and
Fonnat of inf01mat ion which can travel from one pole to another: openness and
suitabi l i ty to the d iversity of stakeholders' competencies.
10.3 A Diversity of Settings
In th is sect ion, we describe some examples of part 1c1patory techniques and
approaches associated with social s imulat ion models . Settings described in this
overview stem from various fields and disc ip l ines. Most of these have already
produced some reviews on part icipatory approaches. For the purpose of the discus
sion in relation with social s imulat ion, a synthesis of these reviews is provided here
with a focus on the requests identi fied in the previous section.
10.3.1 From System Science and Cybernetics
Cybernetics and system sciences have produced a first category of simulation models
of social complexity (Gi lbert and Troitzsch 1999). These models are based on tools
originating from system dynamics, using speci fic software. They focus on flows of
resources and information between stocks which can be control led.
Two main types of i nteractions between these models and stakeholders have so
far emerged: group model bui lding ( Venn ix 1 996); and management fl ight
simulators or microworlds ( Maier and Grossler 2000) .
Group Model Bu i lding experiments focus on the interact ion with stakeholders in
the design stage of a mode l l ing process. l t associates techniques of system dynam
ics mode l l ing with brainstom1ing tools and other techniques of group work, mainly
206 0. Barreteau et al .
based on workshops and meet ings. This trend consists of integrat ing future users of
the model i n the design stage. The part icipants are supposed to be the cl ients of the
model l ing process. Rouwette and colleagues analysed I 07 cases of such
experiments and proposed a number of guide l ines to fac i l i tate consi stent reporting
on part icipatory mode l l ing exercises. These guide l ines focus on three categories:
context, mechanisms and results ( Rouwette et al. 2002) . The second category
focuses predominately on preparation activit ies and description of meetings,
along with factual elements and the mode l ! ing process.
This category of participatory mode l l i ng deals with the expectat ions identi fied in
the f irst section in the fol lowing manner:
The part ic ipation arena is constituted of a rather small or medium size well
ident i fied group. The structure of the interaction is rather global : debates tackle
the whole mode l , and part ic ipants are supposed to be concerned by the model
entity as a whole. The connections may convey information on the tac i t knowl
edge of stakeholders, as we l l as on their purposes. This is st i l l very diverse
among the experiments. The group of stakeholders is mobi l i sed within speci fic
events, workshops, which m ight be repeated. The aim is to feed the model but
also to increase the probabil ity of use of the models produced.
- The process is predominately control led by the model lers; and
- The forn1at of i n formation is general ly not wel l formal ised, even though
techniques, such as hexagons brainstonning or causal diagrams ( Akkermans
1 995 ), appear to organ ise the knowledge brought by stakeholders. Th is low
formal i sation allows the issues related to stakeholder diversity to be tackled
and alleviated in the problem framing phase, but i t leaves a large place to the
model lers' interpretat ion.
Management flight simu lators or microworlds const itute a complementary tech
n ique, which focuses more on the stages of use and simu lation outcomes analysis,
even though this technique may also be used in a design stage to e l ic i t tac it knowledge. A key characteristic of this type of technique is to encourage learning
by doing. Part ic ipants. who might be the c l ients or other concerned people without
any formal rel at ion to the mode l l ing team, have to play through a simulation of the
mode l . Mart in and col leagues have used this techn ique to val idate a system
dynamics model on the hen industry ( Mart in et al. 2007 ) . Part icipants were asked
to play with some parameters of the mode l .
When used to e l ic it knowledge, microworlds attempt to provide events that are s imi lar to those that part icipants already face or are l i ke ly to face in their act ivit ies
related to the issue at stake in the mode l . Le Bars and col leagues have thus
developed a game sett ing to lead farmers to understand the dynamics of their
territory with regard to water use and changes in EU Common Agricultural Pol icy
( Le Bars et a l . 2004 ). In flight simulator experiments, interaction between
stakeholders and the s imulation model is structured around future users of the model or people whose stakes are represented in the model , with a s l ightly deeper
connection than with previous group model l ing bui lding approaches. Part icipants
are asked to deal with parameters of the model and are framed in the categories used
I 0 Part icipatory Approaches 207
in the mode l . There is no a priori di fferentiat ion among part tc tpants. The
connect ions convey infom1ation about the object from the model to part ic ipants.
l t also conveys the part ic ipants' reactions to this object, and some behavioural
patterns observed that can provide new information for the mode llers. This connec
tion is activated by t he part ic ipants working through speci fic events and focus on the use of the tool . Control is sti l l on the side of mode l lers, who frame the
interactions. The format of information is largely fom1ali sed from model to
stakeholders. I t i s not fonnal i sed from stakeholders to mode l .
10.3.2 Knowledge Engineering: Between Artificial Intelligence
and Social Psychology
Knowledge engineering focuses on a specific t ime of the interaction between
stakeholders and a simulation model in the design stage: the process of translating
tac it knowledge into conceptual or somet imes computational models . Many knowl
edge el icitat ion techniques are useful in transfonning written or oral text into pieces
of simu lation models . The purpose of these techniques is to separate the
contributions made directly to the model from the design of the model itself.
Knowledge engineering aims to provide interfaces for th is gap.
To deal with th is interface , techniques have been developed, grounded in art i fi
cial inte l l igence, ( soc ial ) psychology and cognitive science. Behavioural patterns in
social s imulation models are often botTowed in simpli fied versions from these fields
( Moss et al. 2000; Pah l-Wostl NO Hare 2004) . This cross-pol l ination of discipl ines
can be potential ly frui tful for model design. As an example, Abe l and col leagues
have built upon the concept of a mental mode l . They assume that individuals have
representations of their world which may be fonnal ised in causal rules. Working in
the Australian bush, they have designed specific individual interview protocols and
analysis frameworks to e l ic i t these mental models ( Abel et a l . 1 998) . In this case,
interact ion with the model occurs through the interviewer who in this case was also
the mode l ler. There was no col lective interact ion. Researchers deal ing with the
interviews and the corresponding model design clearly guide the process. The
format of information is speech ( i n the fonn of a transcribed text ) , which is
transfom1ed into a mode l l ing language in th is el icitation process.
Bu i lding upon A bel ' s work, Becu has further minimised the involvement of the
model ler, st i l l using individual interviews. This has led him to collaborate with an
anthropologist and to use ethnographic data as a benchmark. I nd ividual interviews,
with the interviewee in the environment su itable to the purpose of the interview, led
him to identi fy objects and relations among these objects. These constitute the
ini t ial basis for an exercise, labe l led as playable stories: stakeholders, in his case
farmers from Northern Thai land, are asked to choose the key elements to describe
their world from their own viewpoints ( w ith the possibi l ity of adding new
e lements) , then to draw relat ions among them and to te l l a story with th is support
208 0. Barreteau et al .
( Becu 2006; Becu et al . 2006) . I n th is case, interaction between stakeholders and the
simulat ion model i s st i l l on an indiv idual basis . The format of conveyed informa
t ion is finally less formal, but the work of translat ion is less important. However,
control of the process sti 11 remains largely in the hand of the modeller, but to a lesser
degree than in previous examples. This technique was further associated with semi
automatic ontology building procedures by Dray and col leagues in order to gener
ate collect ive representat ions of water management in the atoll of Tarawa
( Dray et al . 2006a).
With inspiration coming simi larly from the domain of ethnography, Bharwani
and colleagues have developed the KNeTS method to e l ic i t knowledge. Apart from
a first stage with a focus group, th is method i s also based on indiv idual interv iews.
As in Becu 's work, interaction occurs in two phases: e l ic i tation through
questionnaires and involvement in the model design at the val idation stage, which
is also considered as a learni ng phase for stakeholders. These authors used an
interact ive decision tree to check with stakeholders whether the output of simulat ion would fit their points of view ( Bharwani 2006) . Control of this process is on the
model ler's side. The stakeholders' interact ion is marginally deeper in the model
than in prev ious examples, since there is a d irect interaction with the model as in
management fl ight simu lator. On the other hand, the ontology which is manipu lated
seems to be poorer, s ince the categories of choices open in the interact ion are rather
reduced. The format of infonnation is open in the fi rst phase and very structured in
the deci sion tree in the second phase. The structuration process used in the
model l ing process occurs outside of the field of interaction with the stakeholders.
On its side, Group Decision Support System design domain is based on a
col lect ive interaction with stakeholders as early as the design stage. These systems tend to be used to address higher level stakeholders. In the method he developed,
ACKA, Hamel organ ised a simu lation exercise with the stakeholders of a poultry company. ln this exerc ise, the part icipants were requested to play the ir own roles in
the company. He constrained the exchanges taken place during the exercise through the use of an e lectronic communicat ion medium so that he could analyse them and
keep track of them later. A l l of the partic ipants' communication was transformed
into graphs and dynamic diagrams ( Hamel and Pinson 2005 ). In th is case, the
fom1at of information was qu ite structured.
10.3.3 From Software Engineering
Close to the art i f icial inte l l igence trend, working l ike Hamel and Pinson on the
design of Agent Based Mode ls, there i s an emerging trend in comput ing science
based on agent based part icipatory simulations (Guyot and Honiden 2006) or
part icipatory agent based design ( Ramanath and Gi lbert 2004). Th is trend focuses
on the development of computer tools, mult i -agent systems, which originate from
software engineering. Guyot proposes the implementation of hybrid agents, with agents in the software control led by real agents, as avatars (Guyot 2006 ). These
I 0 Participatory Approaches
avatars help the players' ut
They can be thought as learr
player and are progressively
hybrid agents implement a
simulat ion model. lnformati
assumpt ions, as well as to t�
Ramanath and Gilbert hav
which may be coupled to part
union between software desi
duction not only between d(
with stakeholders contribul
Supported Cooperative Wor
their participation tends to in
mode.
The implementation of it
software development proce
S i lver 1 995 ) al lows issues n
phase, attribut ing a champi
i ssues. This protocol might
may also increase the compL
This process is based on the
Joint application design i :
with a second technique: r through a software developr
or parts of the targeted pr
cri t ic ised, respeci fied, or tt
the final stages of the proc
assessment of the product. T
targeted product .
In these cases, control c
fac i l itator. Otherwise, cont
content of the interaction is
according to part icipants' l i
application design experimE
process is actually rather po'
i s hardly compatible with th
users, compared to the time
rather superficial and needs
speci fic tasks gives a l ittle t the content of pieces of the
Besides these approache
ing in thematic fields sue
workshops that focus on th
are a type of focus group, c
dynamics and interactions,
I 0 Participatory Approaches 209
avatars help the players' understanding the system (Guyot and Honiden 2006) .
They can be thought as learning agents: they learn from choices of their associated
player and are progressively designed ( Rouchier 2003) . The approaches working on
hybrid agents implement a deep connection between part ic ipants and the social
simulat ion model . information conveyed in the interaction i s relat ive to the model
assumptions, as well as to the model content.
Ramanath and Gi lbert have reviewed a number of software engineeting techniques
which may be coupled to participatory approaches ( Ramanath and Gilbert 2004). This
union between software design and participatory approaches is based on joint pro
duction not only between developers but also with end-users. Not only i nteraction
with stakeholders contributes to better software ergonomics - the Computer Supported Cooperative Work (CSCW) workshops series being an example - but
their participation tends to improve their acceptation and further appropriation of the
mode.
The implementation of interactive techniques may take place at all stages of a
software development process. l n early stages, joint appl ication design ( Wood and
Si lver I 995 ) al lows issues raised to be dealt with during the software development
phase, attribut ing a champion to each i ssue. l t is also concerned with technical
i ssues. This protocol might involve other developers, as well as potential users. It
may also increase the computing l iteracy of the participants i nvolved in the process.
This process is based on the implementation of rather well framed workshops.
Joint application des ign is supported by using prototypes. lt is here we find a l ink
with a second technique: prototyping. This technique can be used al l the way
through a software development cycle. l t is based around prov iding rough versions
or parts of the targeted product. For example, i t al lows the pre-product to be
crit ic i sed, respeci fied, or the i nterface improved. Quite c lose to prototyping, in
the f inal stages of the process, user panels can be used to involve end-users in
assessment of the product. These panels are based on a demonstration or a test of the
targeted product.
I n these cases, control of the process is dependent on the h iring of a ski lful
faci l i tator. Otherwi se, control of the process may become rather implicit . The
content of the interaction i s rather technical , which makes i t potential ly unbalanced
according to part ic ipants ' l i teracy in computer science. An assessment of 37 joint
appl ication design experiments has shown that the part icipation of users during the
process is actually rather poor, notably due to the technical nature of debates, which
i s hardly compatible with the t ime al located to a joint appl ication design process by
users, compared to the t ime al located by deve lopers ( Davidson 1 999) . Interaction is
rather superficial and needs translation. However, ident i fication of a champion of
specific tasks gives a l i tt le bit more control to partic ipants, as does involvement in
the content of pieces of the tool bei ng developed.
Besides these approaches originating from soft ware engineering, people work
ing in themat ic fields such as the env ironmental sciences propose eo-design
workshops that focus on the development of s imulation models . Such workshops
are a type of focus group, organised around the ident i fication of actors, resources,
dynamics and interactions, sui table for a set of stakeholders to represent from a
2 1 0 0. Barreteau et al.
socio-ecological system on which they express their own point of view ( Et ienne
2006). Th is approach, which occurs at the design stage of the mode l l ing process is
supposed to lead part icipants to design the simulation model by themse lves, by
formal i si ng the conceptual model through a series of diagrams and a set of logical
sentences. The final interaction diagram and the attached logical sentences are then
translated by the model ler in computer code. I t i s in this type of process that a deep
interaction can occur between part ic ipants and the model . Th is interaction conveys information on the model content, which i s attached to the representations and
knowledge of each part icipant.
10.3.4 From Statistical Modelling
Bayesian Bel ief Networks have been developed to inc lude in the computation of
probabil ities, their dependence on the occurrence of any event. They can be useful to represent complex systems and increasingly used in part ic ipatory settings
because their graph ical nature faci l itates discussion ( Henriksen et al. 2004) . A
group of part icipants can be asked individual ly or col lectively to generate relations
between events and possibly probabi l i t ies as wel l . Henriksen and his col leagues propose a method in seven stages which alternates between individual and col lec
tive assessment and rev ision of an ex istent Bayesian Be l ief Network diagram. This approach is reported to st i l l present some diff icult ies in encouraging strong
pa11icipant involvement due to the mathematical functions beh ind the network
structure. However, other researchers and practit ioners have improved their com
munication and faci l itation of the technique with their own Bayesian Bel ief Net
work processes and are rece iv ing positive stakeholder engagement in the model l ing
processes (Ticehurst et al. 2005 ) . In the example of Henriksen and col leagues, the process i s controlled by the mode l ler and includes only a rather superficial coupl ing
between part ic ipants and the mode l . The translation of part icipant-prov ided infor
mation into probabi l i t ies i s mediated by the model ler and is rather opaque, as in many part icipatory mode l l ing approaches.
10.3.5 From the Social Sciences
The association of part ic ipatory approaches and social s imulation model l ing also
originates from discipl ines not focussing on the production of tools but on under
standing soc ial systems. Social psychology, economics, management and policy
sciences have all developed their own interactive protocols to involve stakeholders
in the design and/or use of their models. Sociology is s t i l l at the beginning of this process ( Nancarrow 2005 ) . These protocols propose a variety of structures of
experimental settings, from laboratory to in v ivo experiments through interactive
platforms ( Cal lon and Muniesa 2006). These three categories vary according to
I 0 Part icipatory Approaches 2 1 1
their openness to the in fl uence given to pat1 icipants. The in v ivo category is beyond
the scope of this paper since it does not involve model l ing: the society in which the
experiment is embedded provides its own model ( Cal lon and Mun iesa 2006) .
Laboratory sett ings are very controlled experiments, i nvolving human subjects.
This is the case for most economic experiments. Pat1 ic ipants are encouraged to
behave with a given rat ionality through instructions and payments at the end of the
session. In canonical experiments, analysis of the experiments is performed by the
scient ist. The focus of the analysis is to understand the indiv idual and collective
behavioural patterns generated by these sett ings. The purpose of these experiments
is either: the test ing of theories and mode ls; new knowledge on human behavioural
patterns in given situations; or the test ing of new institut ional configurat ions
( Friedman and Sunder 1 994). These experiments are part icularly efficient for
situations with strong communication issues or with important inter-individual
interactivi ty (Ostrom et al . 1 994). The issue of simulat ing a real si tuation is not
considered, but rather the test ing of a theoret ical model . This field is currently very active and evolve with the emergence of field experiments involving stakeholders
concerned by the issues ideal ised in the model tested, ask ing them to play in their
env ironment (Cardenas et al . 2000). With this configurat ion, interactions are rather
deep since part ic ipants act as parts of the mode l . The part icipants convey action
choices. However, the experimental ist strongly controls the process.
A platform is an intermediary setting more open to compromise and hybridisation than the laboratory . Heterogeneity of participants is also more welcome, since the
setting is designed to enhance sharing interests. Through experimentation, a platfom1
is supposed to bridge through experimentation the gap between the world of the
model and that of the stakeholders (Cal lon and M uniesa 2006). Policy exercises and
role playing games, as developed in the companion model l ing approach, are kinds of
these platforms ( Richard and Barreteau 2006). Policy exercises embed stakeholders
in potential situations they m ight have to face in the future (Toth 1 988) . They stem
from war games that have been developed since the t ime of Ancient China and are
now used in public pol icy as ·essment ( Duke and Geurts 2004) or environmental
foresighting ( Mermet 1 993) . They are actually quite simi lar to the business games and
the system dynamics trend explained previously in subsect ion I 0.2. 1 . However, the
underlying social simulation model is rather implicit ; though it exists to create the
potential situation and to help ident i fy the participants relevant to the exercise.
Association with a computer tool tends to be with a simulat ion model of the environ
ment, that does not necessary involve a soc ial component. The interaction between
pat1icipants and the social model is rather deep since they are pieces of the model and
connect with the model of their env ironment. Control of the process is rather diffuse.
There might be a genuine empowerment of pa.11icipants since they have the possibi l ity
of bringing their own pa.I1s of the social model to the process, and can adapt i t in ways
different to what the designers expected. A l ike with laboratory settings, platforms
provide infom1ation to the model ler about behavioural patterns of the participants.
Reaction to taboos or innovative behaviours in situations new to the pa.I1icipants, tacit
rout ines, and collect ive behavioural patterns can be el ic i ted using these platforms, whi le it is difficult with classical interv iewing techniques.
2 1 2 0. Barreteau et a l .
Between experimental laboratory settings and policy exercises, the companion
model l ing approach proposes an assoc iation of role playing games and agent based simulations ( Bousquet et al . 2002) . Even though authors in th is approach claim not
to l imit themselves to these two categories of tools, they predominately rest in the
trend of part ic ipatory agent based simulat ions, and are thus close to the software
design and art i ficial intel l igence trends presented above. This approach makes a ful l
use of simi larit ies in archi tecture between role playing games and agent based
simulat ions ( Barreteau 2003 ) . Both implement autonomous agents that interact
within a shared dynamic environment. Joint use of both agent based simulation
and role playing games builds upon these s imilarit ies to express the same concep
tual model . Authors in th is approach use this to reinforce a principle of making al l the assumptions underly ing a model used or design interact ively with stakeholders
explicit and understood. At the design stage, th is approach aims to incorporate
stakeholders' v iewpoints in the model . At the model use stage, it aims to improve
the appropriation of the tool produced, as well as to increase its legit imacy for further operat ional use. However, this appropriation is st i l l under discussion and
might be rather heterogeneous ( Barreteau et a l . 2005 ) .
10.4 Participation in the Modelling Process:
Diversity of Phases and Intensity
Wh.i le many authors c laim to use part icipatory approaches for the simu lation of social complexity, there remains a large diversity of actual involvement of stakeholders and
of act ivit ies h idden behind this involvement. Associations of participatory methods
with soc ial simulation models are rather heterogeneous. lt is thus important to qualify
the actual involvement of stakeholders in these processes. This level of participation can range from mere information received by concerned parties related to the output of a process to the ful l involvement of a wide range of stakeholders at all stages of a
process. There are also many intennediary si tuations imaginable. Pat1icipation should
not be thought of as just talking, and diversity should be made explicit so that criticisms towards part ic ipation as a global category ( lrv in and Stansbury 2004) can
focus on speci fic implementations. This section explores the potential consequences
of this diversity in three dimensions: stage in the model l ing process, degree of
involvement and heterogeneity of stakeholders involved.
10.4.1 Stages in the Modelling Process
The mode l l ing process can be subdiv ided into the fol lowing stages, with the
possibi l i ty of i terating along them:
- Prel iminary synthesis/diagnosis ( through prev iously available data) . Th is
incl udes making expl icit the goal of the model l ing process
- Data collection ( speci fic to the modell ing purpose )
1 0 Part icipatory Approaches
Conceptual model design
Implementation
Calibration and veri ficat ion
2 1 3
Simulation process ( might be running a computer s imulation model, play ing a
game session, etc . )
Validation
Discussion of resu lts
involvement of stakeholders in each of the different stages of the mode l l ing
process does not generate the same level of empowetment or learning, even i f we
assume that th is involvement is sincere. Pre l iminary synthesis, conceptual model
design, val idation and, to some extent, discussion of results are frami ng stages;
stakeholder involvement at these leve ls gives power to stakeholders to orientate the
process. In the pre l iminary synthesis/diagnosis, stakeholders have the opportunity
to play a part in sett ing the agenda. Th is i s the stage of problem structuring which is
identified as a key one in a l l part ic ipatory processes ( Daniel l et al . 2006) . Even if the agenda developed with stakeholder i nvolvement might further evolve, i ts
init ial i sat ion generates a strong i tTeversibi l ity in the process: data col lection,
part icipants select ion and (part ia l ly) mode l l ing choices ( architecture, platform )
are related to this agenda and are costly, e i ther directly or through the necessity of re-programming . The model l i ng process is a sequential decision process, and as
shown in theory of sequential decisions: in i t ia l decis ions are often at the source of
more consequences than envisaged (Henry 1 974; R ichard and Trometter 200 I ) .
Conceptual model design constitutes a landmark in the process. It is the
crystal l i sation of v iewpoints that serves as a reference in further stages. Validation
is the compulsory stage where stakeholders wi l l have the opportunity to check the e ffectiveness of the computer model in representing correctly their behav iours and
ways of act ing. Discussion of results may also const i tute a framing phase, according
to the purpose of the discussion. I f dimensions of discussion are to be defined and model is open to be modified, there is some place for part ic ipants to ( re- )orientate
the model l ing process. Otherwise, if the discussion of results aims to choose from a
few scenarios for example, the choice is very narrow and might be completely
manipu lated. In this regard, it has been shown that for any vote among composite
baskets, it is possible to maintain that one i tem always selected according to the way
the baskets are consti tuted ( Marengo and Pasquali 2003) . A scenario in th is case is a
kind of composite basket .
In other stages of a mode l l ing process, the influence of stakeholder involvement
on the overal l process is less i mportant. When data collection, or cal ibration and
verification involve pat1icipants, stakeholders tend to take the role of informants.
Among the various levels proposed in the classical ladder of part icipat ion explained
in the fol lowing subsection, these stages deal predominately with consultation.
Their involvement i s framed by the format of i n formation which is expected, and on the parts of the model which are to be cal ibrated or val idated. I f the process is
open to modi fication in these frames, the level of part ic ipation might be higher, but
st i l l with a l imited scope.
2 1 4 0. Barreteau et a l .
Implementation stage i s another mean to empower part ic ipants. I t is often
impl icit ly framing. But empowerment through involv ing stakeholders in th is technical act iv i ty is rather to raise their l i teracy in this part and raise the probabil i ty of
their appropriation of the model . Simulat ion stage is basically providing informa
t ion to stakeholders on what is being done. Th is is a technical stage ( running the
simulat ion ) which keeps a part of strategic choices (design of scenarios and
indicators to track the s imulat ion progress ) . I nvolvement of stakeholders in the
technical pa11, such as through role play ing games, increases their knowledge of the
model from inside. provided stakeholders have the l i teracy for that. I nvolvement in
strategic part is connected to the in i t ial stage which has set the agenda. The further
this initial isation has gone in fonnal is ing the quest ions, the less empowering is this
involvement .
10.4.2 Level of Involvement
Level of involvement is a more classical dimension. It is inspired by the classical
h ierarchy of part ic ipat ion levels proposed f irst by Arnstein ( 1 969). Several rev iews
and adaptations have been made since then, with the same focus on power i ssues
( Mostert 2006; Asselt et al. 200 I ) . These works focus on what part ic ipat ion means in decision making terms (the bases of many pol i t ical or democratic theories), with
Democracy Cube ( Fung 2006) or the work of Pateman ( 1 990) and Rocha ( 1 997) . In most of these examples, the emphasis is placed on who (c i t izens, managers or
pol icy makers) has the balance of power for final deci sion-making ( i .e . the choice
phase of a decision process ( S imon 1 977 ) ) but other issues of process are not
spec i fical ly mentioned. Such part ic ipation classi ficat ions, although useful in a
very general sense for the question of part icipat ion in mode ll ing processes, do not
explicit ly treat the i ssue of the place of a model ler or researchers with expert knowledge ( Daniel l et al . 2006).
On these bases, we consider here the f ive following levels in which there are at
least some interact ions between a group of ci t izens and a group of dec ision makers:
- Information supply : cit izens are provided access to information. Th is is not
genuine participat ion s ince it is a one way interaction;
Consultation: sol icitat ion of cit izens ' v iews;
eo-thinking: real d iscussions between both groups;
Co-design: c i t izens have an act ive contribut ion in pol icy design; and
Co-dec ision making: deci sions are taken jointly by members of both groups.
S ince a model l ing process is a kind of dec ision process, this h ierarchy might apply
to model l ing process as wel l . This is a l ittle bit more complicated because two
processe are behind the modell i ng process and the network of interactions cannot
be represented with a group of cit izens and a group of dec ision or policy makers only.
A model l i ng process with the purpose of s imulat ion has two dimensions along
which these scales might be assessed: model content and bui ld ing on one hand; and
1 0 Part icipatory Approaches 2 1 5
control over model use on the other. Though these two dimensions are related, i t is
useful to consider them separately as they provide power and knowledge : either
within the process; or in the system in which the process takes place . Each of these
dimensions is more closely related to speci fic stages in the model l ing process
presented in the previous subsect ion. However, some stages, such as model design
or implementation. contribute to both dimensions.
Therefore we consider the fol lowing categories:
lnf01mation on a mode l ' s content and no control over model use;
Consultation and no control over model use;
D ialogue with model lers and no control over model use;
Dialogue with model lers and control over model use;
Co-bu i ld ing of a model and no control over model use; and
Co-bui lding of a model and control over mode l use;
Each category i s described in the fol lowing sub-sect ion by a flow of interact ions
within an interaction network based on four poles : A, R, M, and P. A stands for al l
people who are involved i n and/or concerned by the social complexity at stake in
the model l ing process. This includes pol icy makers and ci t izens. R stands for
researchers involved in the model l ing process. M stands for the model . P stands
for pol icy makers. P is a subset of A, which gathers the actors who might use
the model and its output for the design of new regulat ions or pol icies concerning
the system as a whole. We chose to gather cit izens and pol icy makers in A, as in the
mode l l ing process they are rather equivalent in their interactions with the
researchers about the mode l . Their dist inction is useful for the second dimension:
model use and dissemination. We assume that the default situat ion is an access of P
members to the output of the model l ing process.
10.4.2.1 Information and No Control
A - R
l P - M
Part ic ipants are informed about the mode l ' s content and the simulat ion by
researchers, who are the only designers. No control over the mode l ' s use or
di ssemination is deputed to part ic ipants as such. Whatever the use of the model
may be afterwards, ci t izens become only better aware of the basis on which this
model has been bui l t . However, the model ex ist s and can be used by members of P.
This i s the classical s i tuation with simulation demonstration and explanat ion of a
mode l ' s assumpt ions. This explanation might be achieved by more act ive means,
such as a role playing game. A switch to the fol lowing category occurs when this
explanat ion leads to a debate that makes the model open to modi fications. Other
wise . it remains mere informat ion.
2 1 6 0. Barreteau et al .
1 0.4.2.2 Consultation and No Control
A - R
! P - M
Part icipants are consulted about the mode l 's content and its s imulat ion that i s by the
researchers, who are the only designers. They prov ide inf01mation and sol ici t
comments on the mode l . Mere data col lection through a survey does not fal l in
th i s category because it assumes act ive involvement from participants in providing
information to the model lers. Some knowledge e l icitation techniques, such as BBN
design, tend to fal l mostly in th is category. Translation of the inputs originating
from participants into pieces of a model is perf01med only by researchers. Th is
translation i s not necessary transparent. No control over use or dissemination of the
model is deputed to part ic ipants as such. Compared to previous category,
patticipants have the abi l i ty to frame marginal ly more of what is performed by
the model through their i nputs to t he mode l ' s content However, the extent of this
abi l i ty depends on the part icipants' ski l ls to ident i fy potential uses of a mode l . As in
any participatory process, when there is an unbalanced power relation between
part ies, the process is also a way for pol icy makers to gain i nformation from
stakeholders; i nformation that could be used for strategic purposes . This bias can
be al leviated if the involvement of A includes all members of A, including the
subset P. The constructed model in th is case may be used by the members of P.
1 0.4.2.3 Dialogue with Modellers and No Control
A - R
� ! p -M
I n this category, iterative and genu inely interactive processes between stakeholders and modellers start to appear. There is sti l l a translation of inputs from part icipants
into the model through the researchers, but there is feedback about these
developments to the stakeholders. Th is leads to discussion about the mode l . Conver
gence of the d iscussion remains on the researchers' side. Group Model Bui lding
experiments predominately fal l into this category. In this case, stakeholders may
increase their influence on the framing of the model with better prior assessment of
the scope of simulations to be examined. B i ases related to strategic information being
revealed in the d ialogue process are sti l l present if there is unbalanced involvement of
Members of A, and notably if members of P are less active, but sti l l present. However,
this category st i l l represents indirect control and no speci fication of model use is left
1 0 Part icipatory Approaches 2 1 7
open to the stakeholders. At the end of the process, the created model can be used by
members of P without any control or any roadmap set by other members of A .
1 0.4.2.4 Dialogue with Modellers and Control
A - R
��! P M
This category is the same as the previous one with translation of stakeholders'
inputs and feedback from the researchers about them. However, the output of the
discussion, the model , is appropriated by stakeholders. They have control over i ts use and dissemination of models which may have been produced through the
mode l l ing process: who might use them; with which protocol ; and what is the
val ue of their outputs. They can decide whether the model and simulat ions are
legit imate to be used for the design of pol icies that may concern them. However.
th is appropriation raises i ssues of d ialogue between researchers and stakeholders
about the su itabi l i ty of model for various uses. Comparison of several part icipatory
agent based simulations has shown that there is a need for dia logue about not only a
mode l ' s content but also on i ts domain of validity ( Barreteau et al . 2005 ) .
1 0.4.2.5 eo- building of a Model and No Control
A
� r P - M
A further stage of empowern1ent of stakeholders through participation in a model l ing
process i s their eo-buildi ng of the model . The design and/or implementation of such a
model are joint act ivit ies between the researchers and stakeholders. Co-design
workshops or joint application development fal l into this category, provided that
there is genuinely no translation of stakeholders' inputs by the researchers.
Techniques originating from Art i ficial Intel l igence and knowledge engineering, as
presented above, aim to reach this leve l, either through the implementation of v irtual
agents extending stakeholders, or through constraining the interactions between
actors through a computer network. This involvement increases the f ide l i ty of the
model to match stakeholders' v iewpoints and behavioural patterns. However, at the
end of the process, the created model can sti l l be used by members of P without any
control or any roadmap set by other members of A .
2 1 8 0 . Barreteau et al .
1 0.4.2.6 Co-building of a Model and Control
A R
��! p M
This category is the same as the previous one, but actors now have control over use
and dissemination of models which may be produced through the process. This
leads to possible stakeholder appropriation of the models, rais ing the same issues as
in Sect. 3 . 2 .4.
10.4.3 Heterogeneity of Actors
Eversole points out the need for part ic ipatory processes to take into account the
complex ity of the society involved including: power relations; inst i tut ions; and the
diversity of v iewpoints ( Eversole 2003) . This is all the more true when appl ied to
the pa11 ic ipatory process of social simu lat ion model I in g. Most sett ings presented in
Sect . 1 0.2 have a l imited capacity to involve a large numbers of people in
interactions with a given version of a model . When interactions convey v iewpoints
or behavioural patterns, heterogeneity may not appear if no attention is paid to i t .
Due to l imits in terms of number of part icipants, part icipatory approaches that deal
with social s imulation mode l l ing involve usual ly representatives or spokespeople.
The issue of their stat ist ical representat iveness i s left aside here, as the aim is to
comprehend the diversity of possible v iewpoints and behav ioural patterns. There i s
st i l l an issue of their representat iveness through their legit imacy to speak for the
group they represent, as wel l as their competency to do so. The feedback of these
spokespersons to their group should also be questioned. When i ssues of empower
ment are brought to the fore, the potential for framing or control l i ng the process is
dedicated to the participants. This might induce echoes in power re lations within
the group, notably due to train ing that may be induced.
Van Daalen and Bots have proposed a categori sation of part ic ipatory model l ing according to th is dimension with three scales: indiv idual involvement; a group
considered as homogeneous, and a heterogeneous group ( Daalen and Bots 2006) .
Table I 0. 1 provides examples of each level according to the two processes involved
that were explained in previous subsection.
These three categories are represented in the diagrams be low, as expansions of
the relation between A and (M U R) in the prev ious subsection. The third category
corresponds to the deep connection mentioned in the first section ( Figs. I 0. 1 , I 0.2,
and 1 0.3 ) .
Some other ways are currently explored with hybrid agents to technical ly over
come the difficulty of deal ing with representat ives: by involving them al l in large
systems. The intemet or mobile phone networks provide the technical substrate for
1 0 Part ic ipatory Approaches 2 1 9
Table 10 . 1 Categories o f part ic ipation according t o level o f heterogeneity embraced ( From Daalen and Bots 2006)
Level Model construction Model use
Computer model Gaming simulation
I . Individual Knowledge e l icitation Model can be executed Indiv idual can ' play ' stakeholders involving one or and indiv idual an actor in a tlight
more individuals stakeholders are simulator setting separately; infom1ed of the (e.g. Maier and depending on the result (e.g. Dudley Grossler 2000: model l ing method 2003) Sterman 1 992) this may consist of interviews about ( perceptions on) a system or questionnaires related to the aspects being model led (e.g. Molin 2005 )
2. Homogenous Same as I , but group Use of a model in a Mul ti-player gaming group model building homogenous group simulation can be
includes interaction means that the model conducted. the between can be run in a game is followed by stakeholders (e.g. workshop setting and a debriefing (e.g. Caste l la et al . 2005 ) model results are Mayer et al . 2005 )
discussed (e.g. Daalen et al . 1 998)
3 . Heterogeneous Same as 2, but group Same as 2, but results Same as 2, but fu l l group model bui lding discussed with stakeholder group
interaction between stakeholders with involved (e.g. stakeholders with di fferent Etienne et al. 2003) d ifferent percept ions/beliefs perceptions/beliefs (e.g. Van den Belt 2004)
such interaction. A large number of participants have a v irtual component in a large
system, interacting with other components, possibly with the purpose of building a
model ( Kiopfer et al . 2004) . However, in this case it is rather an individual interaction
of these participants with the system, than genuine interactions amongst the
participants.
10.4.4 Which Configurations Can Meet the Expectations
of the First Section?
In this subsection we revis i t the expectat ions towards the joint use of partici patory
approaches and social s imulat ion presented in the first section, through the
CIRAD-DIST Unite bibliotheque Lavalette
220
Fig. 1 0. 1 Individual involvement
Fig. 1 0.2 Homogenous group involvement
Fig. 1 0.3 Heterogenous group involvement
�
�
�
0. Barreteau et al .
� I<E-----------?0 �
�
�
I 0 Part ic ipatory Approaches 22 1
Table 1 0.2 Matching expectations on joint use of participatory approaches and social simulation model l ing with categories of participation
Expectation
Increase model 's quality with social diversity and capacity to evolve
Increase mode l ' s quality through distribution of control
Improve suitabil ity of simulation model 's use for increasing knowledge
I mprove suitabi l i ty of simulation model " s use for policy making
Simulat ion as a means to support participation to deal with dynamics and uncertainties
Simulation as a means to support par1 icipation through social learning
Key stage(s) for part icipation
Simulation
Simulation
Design
Design and discussion of results
Discussion of results
Pre l iminary diagnosis, design and discussion of results
M inimum level of empowerment
I n formation and no control
In formation and no control
Dialogue and no control
Dialogue and control
Consultation and no control (depend on partici patory process to be supported)
eo-bui lding and control ( to be preferred)
Level of heterogeneity
Heterogeneous group
Heterogeneous group
Individual
Homogeneous group
Homogeneous group
Heterogeneous group
categorisations above. This is a tentative mapping of part 1c1patory approach
categori sation with model expectations. Table 10.2 be low synthesises this mapping.
The two expectations deal ing with increasing a mode l ' s quality often actual ly
use participants as ( sometimes cheap) resources in the simulation mode l l ing pro
cess. The most important stage is simulat ion, because part ic ipants are supposed to
bring missing information to the s imulation, as wel l as the missing complexity. The
minimum level of empowem1ent is rather low. These processes are hardly part ici
patory in that sense, because pa11icipants are not supposed to benefit from the
process, except a potential payment. A higher level of empowerment might increase
the qual i ty of part icipants ' involvement in the process through a deeper concern in
the outcome of the simulation. Final ly , the heterogeneous group level is obviously
to be respected because i t can instil a deep connection between stakeholders and the
model st i l l and concurrently profit from their i nteractions with each other.
To make simulation models match their intended use, the key stage is the design
process. Stakeholders are supposed to aid the bui lding of an appropriate model . The
main difference between targets of simu lation model ' s use is in the necessity to give
control over the process to stakeholders in case of pol icy making. New knowledge is
of i ndividual benefit to al l part icipants and the emergence of fruitful interactions can
also become an indiv idual benefit. There are few direct consequences of this new
knowledge. Therefore, control over the process in this case is useless and involvement
222 0. Barreteau et al.
might be indiv idual, as with knowledge el icitat ion techniques. However, h igher level
of stakeholder heterogene ity might raise the knowledge acquired in the process.
When simulation is used to support participation, discussion of results is a key
stage. Previous stages aid in the problem framing and l iteracy increase of participants
that allow them to reach more sol id interpretations. The empowerment level is rather
dependent on the participatory process that is being suppot1ed. However, consultation
in the model l ing process should be a minimum requirement so that uncet1ainties and
dynamics tackled by the simulations are relevant to the stakeholders. When focusing
on social leaming, eo-bu ilding and control should be preferred because this category
increases the potential for exploration and creativity. However, some social learning
might take place in lesser levels, provided that group heterogeneity is well encouraged in the process.
1 0.5 Combining Approaches and Techniques at Work
We present in this section two case studies implementing various methods for
joining social s imulat ion model l ing and part ic ipatory approaches. The first deals
with fire hazard prevention in southern France, and the second one with groundwa
ter management in the Atoll of K iribat i .
10.5.1 The Fire Hazard Case Study
In December 2005, the Forest Serv ice of the Gard Department of Agriculture
( DDAF), decided to start a f ire prevention campaign focused on f ire hazard at the
interface between urban and forest areas. Interested in the part icipatory approaches
developed by INRA researchers on f ire prevention and forest management planning
( Etienne 2003 ), they ask for an adaptat ion of the SylvoPast model to the periurban
context in order to make local pol it icians aware of the increasing f ire hazard problem. The District of Nlmes City ( N M ) who was already interested in the use
of role playing games for empowering stakeholders and dec ision makers, asked the
Ecodevelopment Un it of the INRA of Avignon to develop a companion model l ing
approach based on social s imulations and a part icipatory involvement of al l the
mayors of the district.
The model l ing process was subd iv ided into seven stages:
I . Collection and connect ion on a GIS of relevant cartographic data on forests, land-use and urban ization, and indiv idual interv iews with local extensionists on
farmers, foresters and propet1y developers practices. 2 . eo-construct ion with DDAF and NM of a virtual but implicit map representing
three v i l lages typical from the northern area of Nlmes city and val idation of the
map ( shape, land-use attributes and scale ) by a group of experts (EX) covering the
0. Barreteau et al .
1iques. However, h igher level
� acquired in the process.
jiscussion of results is a key
teracy increase of participants
; empowerment level is rather
mrted. However, consultation
ment so that uncertaint ies and
stakeholders. When focusing
referred because this category
lowever, some social learnjng
p-oup heterogeneity is well
ques at Work
nenting various methods for
'I approaches. The first deals
e second one with groundwa-
I Department of Agriculture
focused on fire hazard at the
1 the part icipatory approaches I forest management planning
·oPast model to the periurban
1f the increasing f i re hazard
already interested in the use
1d decision makers, asked the relop a companion model l i ng
'atory involvement of all the
stages:
cartographic data on forests,
'IS with local extensionists on
�s.
but implicit map representing
nes city and validation of the
p of experts (EX) covering the
I 0 Part icipatory Approaches 223
main act tv t t tes of the tetTitory (agriculture and l ivestock extension ists. forest
managers, hunting manager, land tenure regulator, fire brigade captain and town
planner).
3. eo-construction, with N M . DDAF and experts of a conceptual model account ing
for the cutTent funct ioning of the territory and the probable dynamic trends to occur during the next 1 5 years. This part icipatory process fol lowed the ARDI
methodology mentioned in Sect. I 0.2.3 ( Et ienne 2006 ).
4. Implementat ion of the NimetFeu model on eormas mult i -agent platforn1 by
INRA researchers and val idat ion of the model by simu lating with the eo
construction group, the current si tuation and its consequences on fire hazard
and landscape dynamics for the 1 5 fol lowing years.
5. eo-construction and test of a role-play ing game ( RPG ) using the N imetFeu
model as a way to s imulate automat ical ly natural processes and some social
decisions (v ineyard abandonment, horse herding, f ire fight ing) . The other social
decisions were programmed to be taken directly by the players and used as an
input to the model .
6. Use of the RPG during several sessions gathering 6 players ( 3 mayors,
I developer, l NM representat ive, I DDAF technician) unt i l the 1 4 v i l lages
involved in the project did part ic ipate to a session.
7 . Adaptation of N imetPasLeFeu to other ecological condit ions, and dec ision of the
Gard Department to become autonomous in running RPG sessions. A faci l itator
and a data manager were trained and tested during sessions organized in the
framework of an I NTEREG project with mayors and fire prevent ion experts from France, Spain , Italy and Portugal
The approach is based on a mutual comprehension of the elements of the tetTitory
that make sense w ith the question asked. This sharing of representations is done by
means of a series of collect ive workshops during which Actors, Resources, Dynamics
and I nteract ions ( ARDl ) which make the stakes of the tetTitory are identi fied and
el icited. To fac i l i tate this sharing, the answers to the questions are fonnal ized into
easily comprehensible diagrams, with a min imum of coding making it possible to
class ify the provided inf01mat ion. The role of the faci l itator only consists in cal l ing
upon each partic ipant, wri t ing down the proposals in a standm·d way, and asking for
reforrnulating when the proposal is too generic, enounced with a polysemous word or
can lend to confusion.
In both models, the env ironment is div ided into three neighbouring vi l lages
covering the gradient of urban ization and agricultural land/woodland ratio currently
observed around Nlmes c ity. I t is visual ized by means of a cel lular automaton through
a spatial grid representing 1 8 land-use types that can change according to natural
transit ions or human activ it ies.
Four categories of social entities are identi f ied: property developers, mayors,
farmers and fire prevention managers. The developers propose new urban
developments according to soc ial demand and land prices. They have to respect the
government regulations ( flood hazard, protected areas, urban zoning). Mayors select
an urbanization strategy (to densify, to develop on fal low land, olive groves or
224 0. Barreteau et al .
forests), update their urban zoning according to urban land avai labi l i ty and social
demand and make agreements with the developers. When updating the urban zoning,
they can create new roads. Fanners crop their fields using or not cun·ent practices that
impact fire hazard (v ineyards weeding, stubble ploughing) or adapt to the economic
crisis of certain commodities by uprooting and set t ing aside lowland v ineyards or
ol ive groves near to urban zones. The fire prevent ion manager establishes a fuel-break
in a strategic place, selected according to f ire hazard ranges in the forest and the
possible connections with croplands, as well as available funds and forest cleaning
costs.
Four biophysical models i ssued from prev ious researches and adjusted to the local
conditions are integrated to the MAS to account for fal low development, shrub
encroachment, pine overspreading and fire propagation. The model is run at a
! -year t ime step, the state represented on the map cotTesponding to the l and cover
at the end of June ( beginning of the wi ldfire period) . Each part icipant was invited to
propose a set of key indicators that permit them to monitor key changes on ecological
or socio-economic aspects. A common agreement was made on what to measure, on
which entity, with which unit , and on the way to represent the cotTesponding qualitat ive or quantitative value ( visualizing probes on graphs or v iewpoints on
maps). They were encouraged to elaborate simple legends, in order to be able to
share their point of v iew with the other participants while running the model.
The first MAS was exclusively used to support the collective thinking on which
procedures and agents wi l l be affected to players, and which ones would be
automat ically simu lated by the computer. In the RPG model , the playing board
was strongly simpl i fied with only four types of land cover. Runn ing the game gives
part icipants the opportunity to play individually or collect ively by turns, according
to a prec isely defi ned sequence. Whi le the mayors players draw the l imits of the
urban zone and rank the price of constructible land according to i ts cutTent land-use,
the developer player sorts randomly a development demand and elaborates a
strategy (v i l l age, density, l i vel ihood) . Then begin a series of negot iations between
the developer and the three mayors in order to decide where to build, at which
density and with which type of fire prevent ion equipment. All the players' decisions
are input into the computer and landscape dynamics is s imulated by running the
model . Players get d ifferent types of output from the simulation run: budget
updat ing, new land-use mapping, popularity scoring. Each round con·esponds to a
3-year lapse and i s repeated three to four t imes according to players ' availabi l ity .
A spec i fic effort i s made in the RPG design to account for physical remoteness
and territory identity among part icipant: the playing room is set up into three
neighbouring but dist inct boxes for the three mayors ( each box represents one
v i l lage ), one isolated smal l table for the developer, and another game place with
two tables, one small for the DDAF and a huge one for N M . Last ly , in a corner, the
computer stuff i s placed with an interactive board than can both be used as a screen
to project different v iewpoints on the map or as an interactive town plan to identify the parcels ' number.
At the end of the game, all the part icipants are gathered in the computer room
and discuss col lect ively, with the support of fast replays of the game played.
10 Part icipatory Approaches 225
Table 10.3 Classification of type of part icipation m various stages of the NimetPasleFeu experiment
Involvement Heterogeneity nb
Prel iminary diagnosis Consultation Individuals 1 0
Data collecting Consultation Indiv iduals 3
Conceptual model designing eo-design Heterogeneous group 1 4 Implementing Information Indiv iduals 2
Calibrat ing and validat ing eo-th inking Heterogeneous group 1 4
Role-playing game designing eo-design Heterogeneous group 1 4
RPG playing and debriefing eo-decision making Heterogeneous group 30
Getting self sufficient lnfom1ation Indiv iduals 3
Di fferent topics are tackled related to ecological processes (effect of f ire , main
dynamics observed) , allitudes (main concerns, changes in practices), and soc ial
behaviours ( negotiations, al l iances, strategies ) .
Along these various stages, th i s experiment features a diversity of involvement
as well as of structure of interactions. Th is is synthesised in the Table I 0.3 above.
10.5.2 The AtollGame Experiment
This study is carried out in the Republic of K iribat i , on the low-lying atol l of
Tarawa. The water resources are predominant ly located in freshwater lenses on the
largest islands of the atol l . South Tarawa is the capital and main population centre
of the Republ ic . The water supply for the urban area of South Tarawa is pumped
from horizontal in f i l tration gal leries in groundwater protection zones. These cur
rently supply about 60 % of the needs of South Tarawa's commun it ies. The
government 's declaration of water reserves over privately owned land has lead to
confl icts, i l legal settlements and vandal ism of public assets ( Perez et al. 2004) .
The AtollGame experiment aims a t providing the relevant information to the local
actors, including inst i tut ional and local community representat ives, in order to fac i l i
tate dialogue and to help dev ise together sustainable and equitable water management
practices. Knowledge e l ic itat ion techniques as well as Mult i Agent-Based
Simulations ( MABS) coupled with a role-playing game have been implemented to
ful f i l this aim. In order to col lect, understand and merge v iewpoints coming from
different stakeholders, the fol lowing five-stage methodology is applied: ( I ) collecting
local and expert knowledge; ( 2 ) blending the di fferent v iewpoints into a game-based
model; ( 3 ) playing the game with the different stakeholders; (4 ) formalising the
different scenarios investigated in computer simulat ions: and ( 5 ) exploring the
simu lated outcomes with the di fferent stakeholders ( Dray et al. 2006b).
In j t ial knowledge el icitation ( Stages I and 2) re l ies on three successive methods.
First, a Global Targeted Appraisal focuses on social group leaders in order to col lect
di fferent standpoints and their art iculated mental mode ls . These col lect ive mode ls
are partly val idated through Indiv idual Act iv it ies Surveys focusing on behavioural
226 0. Barreteau et a l .
patterns of indiv idual islanders. Then, these indiv idual representat ions are merged
into one collective model using qual itat ive analys is techniques. This conceptual
model is further s impli fied in order to create a computer-assisted role-playing game
( Atoll Game) . The range of contrasted v iewpoints confirms the need for an effective
consultat ion, and engagement of the local populat ion in the design of future water
management schemes in order to warrant the long-term susta inabil ity of the system.
Clear evidence of the inherent duality between land and water use rights on one
hand, and between water exploitation and distribution on the other hand, provides
essential features to frame the computer-assisted Role Playing Game.
The assistance of a computer is needed as far as interact ions between groundwa
ter dynamics and surface water balance involve complex spat ial and t ime
dependent interactions ( Perez et al. 2004). The use of Agent-Based Model l ing
( A B M ) enables us to take ful l advantage of the structure of the conceptual model .
We developed the AtoiiGame simulator with the COR MAS© platform ( Bousquet
et al. 1 998) . 1
A board game version reproduces the main features of the Atoll Game simulator
( Dray et al . 2006a). 1 6 players - 8 on each is land - are able to interact according to a
set of pre-defined rules. Their choices and act ions are directly incorporated into the
simulator at the end of each round of the game. During the game, players can ask for
more information from the simulator or discuss the results provided by the simula
tor ( sal in ity index , global demand) . Landowners, tradit ional or new buyers, are the
essential actors in the negotiat ions with the government. The connection between
land tenure issues and water management is essentia l . It drives the land use
restrictions and land leases discussions. The populat ion increase, mainly through
immigrat ion, is perceived as a threat in terms of water consumption, pol lut ion
generation and pressure on the land. Financial i ssues l inked with water management
usual ly deal with land leases, equ ipment investment and, seldom, with water
pricing. Hence, the model features:
Agents/players becoming a local landowners;
Land and water al location confl ict ing rules, and various sources of incomes;
An increasing number of new sett lers on agents/players' land.
The indiv idual object ive of the players is to minimize the number of angry or
sick people in their house. People may become ANGRY because they didn't have
enough water to drink during the round. People may become S ICK if they drank
unhealthy ( pol luted or salty ) water during the round. POLLUTION depends on the
number of people l i ving on the i sland and contaminating the freshwater lens.
SALTY WATER depends on the recharge rate of the fresh water lens and the
location of the people on the is land.
At fi rst, representatives from the di fferent islands displayed di fferent v iewpoints
about the Water Reserves. Hence, the group meetings organized in the v i l lages prior
1 More details about the Atoi iGame can be found onl ine al hnp://cormas.cirad.fr/en/applica/ atoiiGame.hlm.
1 0 Part icipatory Approaches 227
Fig. 10.4 Flowchart of financial. technical, and social solutions agreed on by the participants of the AtoiiGame experiment
the workshop al lowed for a real ly open debate. On the institutional side, the position
of the di fferent officers attending the workshop demonstrated a c lear commitment to
the project . All the participants showed the same level of mot ivation either to
express their v iews on the issue or to genuinely try to understand other viewpoints. Part icipants also accepted to fol low the ru les proposed by the project team, espe
cially the necessity to look at the problem from a broader perspect ive. During the
first rounds, the players quickly handled the game and entered into interpersonal
discussions and comparisons. The atmosphere was good and the game seemed
playful enough to maintain the part icipants ' interest al ive. The second day, the
introduction of a Water Management Agency and the select ion of its ( v it1ual )
Director created a l i ttle tension among the participants. But, after a while, the
players accepted the new situat ion as a gaming scenario and started to interact
with the newly created institution. At this stage, players started to mix arguments
based on the game with other ones com ing directly from the real ity . On Is land I ,
players entered direct negot iations with the (v i rtual ) Director o f the Water Manage
ment Agency. On I sland 2, discussions opposed players wi l l ing or not to pay the fee .
Final ly , the project team introduced the fact that the Water Management Agency
was no longer able to maintain the reticu lated system due to a poor recovery of the
service fees. l l had for immediate consequence a sharp decrease of the water
quantity offered on I s land 2.
Then, players from both tables were asked to l ist solutions to improve the
situation on their is land. When the two l i sts were completed, the project team and
the participants bu i l t a flowchart of f inanc ial , technical and social solutions, taking into account issues from both is lands ( Fig. I 0.4 ) .
228 0. Barreteau et al .
A collective analysis of the flowchat1 concluded that the actual s i tuation was
largely unsustainable e ither from a financ ial or soc ial v iewpoint . The flowchart
above provides a set of inter-dependent solut ions that shou ld be explored in order to
gradual ly address the present s ituation.
10.6 Discussion : Relations Between Participants and Models
The diverse categories of joint implementat ion of part icipatory approaches and
social s imulation mode l l ing feature a diversi ty of relations between a set of people,
participants, and a model .
Classical soc ial simulation models do not feature any part ic ipant. People are
represented in the model , sometimes from assumed or theoret ical behavioural
patterns. Th is entai ls exploring potential emergent phenomena from interact ions
among these behavioural pattems. Some part ic ipatory approaches involve only an
implicit social model . Within th is scope, there is a large diversity of re lations. Th is
diversity is based on the role undertaken by stakeholders, their actual involvement
and issues tackled by the model .
I n a l l the processes al lying social simulation models and participation,
stakeholders take on various roles: pieces in simu lation, i ntetfaces for coupl ing
various sources of knowledge, beneficiaries of the process, key informants . . . As pointed out by Ryan, managers are overwhelmed by the complexity to be managed.
Part icipation is a way to share this burden ( Ryan 2000). Stakeholders provide the
missing interactions and add missing pieces of knowledge, such as tacit knowledge
(Johannessen et al. 200 I ) . I f involvement of stakeholders is useful for principal agents
such as managers, we propose it as a rule that they should gain some empowerment in
the process.
Stakeholders can be key pieces of the mode l l i ng process i tsel f as wel l . ln the
simulat ion they are an al ternative to computer code to prov ide the engine
( Hanneman 1 995 ). They provide an answer to issues of coupl ing several v iewpoints
( Robinson I 99 1 ).
However actual involvement of people in a part ic ipatory model l ing process
might largely di ffer from fonnal involvement planned. Leav ing aside cases of
manipulat ion and announce effect, people have also to find their place in the
part icipatory process. Sui tabi l i ty of part ic ipatory approaches in a specific society
has to be taken into account: context ( i nc luding social) is a key driver for success in
stakeholder involvement ( Kujala 2003) , and pract ice of interactive pol icy making
processes depend on local culture ( Driessen et a l . 200 I ) . Representation
mechan isms have already be pointed out as a major factor. I t has to be tuned to
th is local social and cultural context . At a finer grain, fac i l i tator has a key role to
lead people towards the level of involvement they are invi ted (Akkem1ans and Yennix 1 997) .
I 0 Part icipatory Approaches 229
10.7 Conclusion
This chapter provides a review of the diversity of association of participatory
approaches and social s imulat ion, for the ir mutual benefit. This diversity of
approaches al lows tackl ing expectations about i ncreasing mode l ' s qual i ty, mode l ' s
sui tabi l ity to i t s i ntended use and improv ing partic ipat ion. Their diversity i s bui l t
upon ingredients coming from various discipl ines from soc ial sciences to computer
sciences and management. lt is expressed according to the implementat ion of
i nteract ions between the part icipants and the simulat ion mode l , the control of the
process and the format of information. Thi s leads to expand the c lassical ladder of
part ic ipation towards categorization according to the stage in the model l ing process
when part icipation takes place and the structure of the interaction to cope with the
heterogeneity of stakeholders.
This diversity requires a cautious description of each implementation in s itua
t ion, so that any evaluation is speci fic to the implementat ion of a given assoc iat ion
in i ts context. General i sat ion can then be done only on the relation of this practice of
part ic ipatory simulation and its suitabi l ity to its context and purpose. Efficiency to
induce changes i n practice or knowledge depends on the respect of a triple cont in
gency of collective deci sion processes: t ime, people, means ( Miettinen and
Yirkkunen 2005) . This means to respect and take into account the own dynamics
within the social system at stake, to allow the part ic ipation of people with their
whole essence ( i nc luding tacit knowledge , networks, relations to the world), and to
be adaptive to means and competences present within the system (Baneteau 2007) .
Another dimension of evaluation should be democracy, since i t is often put t o the
front. This raises the issue of t he existence of a control of the process. Does it rely
only on model lers or is i t more shared? Finally there is a necessity of being more
explicit on the k ind of PA which is used because of the potential deconsideration of
the whole fami ly i f expectat ions are deceived.
Further Reading
Partic ipatory model l ing is i ncreasingly present in specia l sessions of conferences or
spec ial features of scient ific journals. A first source of further readings consists in
case studies. Among others, Environmental Mode l l ing and Software had a spec ial
issue on model l ing with stakeholders ( Bousquet and Yoinov 20 I 0) , where readers
wi l l find a whole set of wel l described case studies using various methods. The
biennial i nternational Environmental Mode l l ing and Software conferences have
also specific tracks for part icipatory model l ing; proceedings are available onl ine
( see http://www. iemss.org/society/ under publicat ions ) . For spec ific tools, refer to
the papers of a symposium on simulation and gaming in natural resource manage
ment, publ ished as a special issue of Simulation and Gaming ( volume 38, issues
2 and 3 ) . The introductory paper g iv ing an overv iew is ( Barreteau et al. 2007 ) .
230 0. Barreteau et al .
Reflex iv ity i s cruc ial for practit ioners of participatory processes, as part of the
need for more cautious eval uat ion of participatory processe as pointed out by
Rowe and Frewer ( 2004 ). Another direct ion for reading consists in methods for
evaluat ion and assessment of stakeholder i.nvolvement in model l ing processes.
Etienne edited a whole book aiming at assessing consequences of a specific
approach, so-cal led companion mode l l ing ( Et ienne 20 1 1 ) .
Readers who are more interested in stakeholder i nvolvement in mode ll ing at a
more technical level should go for the review paper of Ramanath and Gi lbert ( 2004)
which provides a nice overv iew of this point of view.
References
A bel N, Ross H , Walker P ( 1 998) Mental models in rangeland research, communication and management. Range! J 20:77-9 1
Akkermans HA ( 1 995) Developing a logistics strategy through participative business model l ing. lnt J Oper Prod Manag I 5 : I 00- 1 1 2
Akkermans HA, Venni x JAM ( 1 997) Cl ients" opinions on group model building: an exploratory study. Syst Dyn Rev 1 3 :3-3 1
Arnstein S ( 1 969) A ladder of cit izen part icipation. J Am Plann Assoc 35: 2 1 6-224 Barreteau 0 ( 2003) The joint use of role-playing games and models regarding negotiation
processes: characterization of associations. J Art if Soc Soc Simul 6(2), http://jasss.soc.surrey. ac.uk/6/2/3.html
Barreteau 0 ( 2007 ) Modeles et processus de decision collective: entre comprehension et faci l itat ion de la gestion concertee de la ressource en eau ( H DR Thesis). Paris Dauphine University. Paris
Barreteau 0, Bousquet F ( 1 999) Jeux de roles et validation de systemes mult i-agents. In: Gleizes M-P, Marcenac P (eds) lngenierie des ystemes mult i-agents, actes des 7
''"'0' J F!ADSMA.
Hermes, Paris, pp 67-80 Barreteau 0, Bousquet F, Attonaty J-M ( 200 I ) Role-playing games for opening the black box of
mult i-agent systems: method and teachings of i ts application to Senegal River Valley irrigated systems. J Art i f Soc Soc Simul 4(2) . http://jasss.soc.surrey.ac.uk/4/2/5 .html
Barreteau 0, Hare M , Krywkow J, Boutet A (2005) Model designed through participatory processes: whose model is it? I n : Ferrand N, Perez P, Batten D (eds) Joint conference on multiagent model l ing for environmental management. CABM-HEMA-SMAGET 2005. Bourg St Maurice - Les Arcs. 2 1 -25 Mar 2005
Barreteau 0, Le Page C, Perez P ( 2007) Contribution of simulation and gaming to natural resource management issues: An introduction. Simul Gaming 38(2) : 1 85- 1 94
Becu N ( 2006) Identification et model i sation des representations des acteurs locaux pour la gestion des bassins versants. PhD thesis, Sciences de ! "eau, Universite Montpe l l ier 2. Montpel l ier, France
Becu N. Barreteau 0. Perez P, Saising J, Sungted S ( 2006) A methodology for identifying and fonnal izing fanners' representations of watershed management: a case study from Northern Thailand. I n : Bousquet F, Trebui l G, Hardy B (eels) Companion model ing and mult i-agent systems for integrated natural resource management in Asia. I R R L Los Baiios. pp 4 1 -62
Bharwani S (2006) Understanding complex behavior and decision making using ethnographic knowledge el icitation tools ( K nETs). Soc Sci Comput Rev 24:78- 1 05
Bousquet F, Voinov A (eels) ( 20 I 0) Thematic issue - model l ing wi th stakeholders. Environ Model I Softw 25( 1 1 ): 1 267- 1 488
I 0 Part icipatory Approaches 23 1
Bousquet F. Bakam I. Proton H.Le Page C ( 1 998) Conm1s: common-pool resources and mult iagent systems. In : Lecture notes in art ificial intell igence 1 4 1 6, Springer, Berl in pp 826--838
Bousquet F , Barreteau 0, Le Page C, Mu l ion C, Weber J ( 1 999) An environmental modell ing approach: the use of mult i -agent s imulations. I n : B lasco F, Weil l A (eds) Advances in env ironmental and ecological model l ing. E lsevier, Amsterdam, pp 1 1 3- 1 22
Bousquet F et al (2002) Mult i-agent systems and role games: an approach for ecosystem eomanagement . I n : Janssen M (eel) Complexity and ecosystem management: the theory and practice of mult i-agent approaches. Edward Elgar, Northampton. pp 248-285
Callon M, Muniesa F (2006) Economic experiments and the construction of markets. In : MacKenzie D, Muniesa F, S iu L (eels) Do econom ists make markets? On the perfom1at ivi ty of economics. Princeton U niversity Press, Princeton, pp 1 63- 1 89
Cardenas J-C. Stranlund J. Wi l l i s C (2000) Local environmental control and institutional crowding-out. World Dev 28: 1 7 1 9- 1 733
Castella JC, Tran Ngoc T, Boissau S ( 2005) Participatory simulation of land-use changes in the northern mountains of V ietnam: the combined use of an agent-based model , a role-playing game, and a geographic information system. Ecol Soc 1 0( I ) : 27. hllp://www.ecologyand society .org/vol l 0/iss I /art27 I
Cockes D. l ve J ( 1 996) Mediat.ion support for forest land al locat ion : the S I RO-MED system. Environ Manage 20( I ) :4 1 -52
Daalen CE van, Bots PWG (2006) Part icipatory model construction and model use in natural resource management. ln: Proceedings of the workshop on fonnalised and non-formalised methods in resource management - knowledge and learning in part icipatory processes, Osnabri.ick, 2 1 -22 Sept 2006, http://www.part izipa.uni-osnabrueck.de/wissAbschl uss.html
Daniel! KA. Ferrand N. Tsoukias A ( 2006) Investigat ing pm1icipatory model l ing processes for group decision aiding in water planning and management. In : Seifert S. Weinhardt C (eels) Proceedings of group decision and negotiation (GDN) 2006. International conference. Karlsruhe, 25-28 J une 2006, Universitatsverlag Karlsruhe, Karlsruhe, pp 207-2 1 0
D ' Aquino P. Le Page C . Bousquet F , Bah A ( 2003) U sing self-designed role-playing games and a mult i-agent system to empower a local decision-making process for land use management: the SelfCom1as experiment in Senegal. J Art if Soc Soc Simul 6(3), hllp://jasss.soc.surrey.ac.uk/6/ 3/5.html
Davidson EJ ( 1 999) Joint application design (JAD) in practice. J Syst Softw 45 : 2 1 5-223 DeSanctis G, Gallupe RB ( 1 987) A foundation for the study of group decision support systems.
Manag Sci 33 :589-609 Dewulf A, Bouwen R, Tai l leu T (2006) The mult i-actor simulation ' Podocarpus National Park' as
a tool for teaching and researching issue framing. In : Proceedings of LACM 2006 Montreal Meetings. hllp://ssm.com/abstract =9 1 5943
Dray A et al (2006a) The AtoliGame experience: from knowledge Engineering to a computerassisted role playing game. J Art if Soc Soc Simul 9( I ). http://jasss.soc.surrey.ac.uk/9/ I /6.html
Dray A , Perez P. Le Page C, D'Aquino P. White I ( 2006b) AtollGame: a companion model l ing experience in the pacific. In: Perez P, Banen D (eels) Complex science for a complex world: exploring human ecosystems with agents. ANU E Press, Canberra, pp 255-280
Driessen PPJ, Glasbergen P, Verdaas C (200 1 ) Interactive policy making: a model of management for publ ic works. Eur J Oper Res 1 28 :322-337
Drogoul A, Vanbergue D, Meurisse T ( 2003) Mult i-agent based simulation: where are the agents? In : Sichman JS, Bousquet F, Davidsson P (eels) Mult i-agent-based simulat ion 1 1 : third international workshop, MABS 2002, Bologna, Revised papers ( Lecture notes in computer science. 258 1 ) , Springer, Berl in/Heidelberg, 1 5- 1 6 Ju ly 2002, pp 1 - 1 5
Dudley R G ( 2003) Model ing the effects o f a log export ban i n Indonesia. Syst Dyn Rev 20:99- 1 1 6 Duke RD. Geurts J L A ( 2004) Policy games for strategic management. Dutch University Press,
Amsterdam Etienne M ( 2006) Companion model l ing: a tool for dialogue and concertation in biosphere
reserves. In : Bouamrane M (eel) B iodiversity and stakeholders: concertation i t ineraries, Biosphere reserves - technical notes I , Unesco, Paris, pp 44-52
232 0. Barreteau et a l .
Etienne M (ed) ( 20 1 1 ) Companion model ing: a participatory approach to support sustainable development. QUAE. Versai l les
Etienne M ( 2003) SYL VOPAST a mult iple target role-playing game to assess negot iation processes in sylvopastoral management planning. Journal of Artificial Societies and Social Simulations 6, hll p://jasss.soc.surrey .ac.uk/6/2/5.html
Etienne M. Le Page C. Cohen M ( 2003 ) A step by step approach to build up land management scenarios based on mult iple viewpoints on mult i-agent systems simulations. J Art i f Soc Soc Simul 6(2), http://jasss.soc.surrey.ac .uk/6/2/2.html
Eversole R ( 2003) Managing the pitfal ls of participatory development: some insight from Austral ia. World Dev 3 1 :78 1 -795
Fischer G. Giaccardi E, Eden H , Sugimot i M, Ye Y ( 2005) Beyond binary choices: integrat ing indiv idual and social creat iv ity. In! J Hum-Comput Stud 63 :482-5 1 2
Friedman D. Sunder S ( 1 994) Experimental methods, a primer for economists. Cambridge University Press. Cambridge
Fung A ( 2006 ) Varieties of part ic ipation in complex governance. Public Adm Rev 66:66-75 Funtowicz SO, Mart inez-Aiier J. Munda G, Ravetz JR ( 1 999) l n fom1ation tools for environmental
policy under conditions of complex i ty, vol 9. Environmental issues series. European Env ironment Agency. Copenhagen
Gi lbert N. Troitzsch KG ( 1 999) Simulation for the social scient ist . Open University Press, Buckingham
Green KC ( 2002) Forecasting decisions in conflict situations: a comparison of game theory, roleplaying and unaided judgment. In! J Forecast 1 8 :32 1 -344
Guyot P ( 2006) S imulations mult i-agents part icipatives. PhD thesis, l n forn1atique. Universite Paris V I , Paris
Guyot P. Honiden S ( 2006) Agent-based part icipatory simulations: merging mult i-agent systems and role-playing games. J Art i f Soc Soc S imul 9(4). hup://jasss.soc.surrey.ac.uk/9/4/8.html
Hamel A. Pinson S ( 2005) Concept ion part icipative de simulations mult i-agents basee sur une approche d 'analyse mult i-acteurs. I n : Drogoul A, Ramal E (eds) Systemes mult i -agents: Vers la conception de systemes artificiels socio-mimetiques ( J FSMA 2005) . Hem1es. Paris. pp 1 - 1 5
Hanneman R A ( 1 995 ) Simulation model ing and theoretical analysis i n sociology. Sociol Perspect 38:457-462
Henriksen HJ . Rasmussen P, Brand! G, von Biilow D, Jensen FV (2004) Engaging stakeholders in construction and validation of Bayesian belief networks for groundwater protection. In: Proceedings of I FAC workshop on model l ing and control for participatory planning and managing water systems, Venice. 29 Sept- 1 Oct 2004
Henry C ( 1 974) Investment decisions under uncertainty: the i rreversibi l i ty effect . Am Econ Rev 64: 1 006- 1 0 1 2
Hochman Z, Hearnshaw H , Barlow R , Ayres J F, Pearson CJ ( 1 995) X-breed: a mult iple domain knowledge based system integrated through a blackboard architecture. Agr Syst 48:243-270
l rvin RA, Stansbury J ( 2004) Cit izen participation in decision making: is i t worth the effort. Public Adm Rev 64:55-65
Johannessen J-A, Olaisen J. Olsen B (200 I ) Mismanagement of tacit knowledge: the importance of tacit knowledge. the danger of information technology, and what to do about i t . In! J l n f M a nag 2 1 :3-20
Kinzig A, Ryan P. Etienne M. Al lyson H, Elmqvist T, Walker B (2006) Resil ience and regime shifts: assessing cascading effects. Ecol Soc 1 1 ( I ) : 20. hllp://www.ecologyandsociety.org/ vol l I /iss I /art20
Klopfer E, Yoon S. R ivas L ( 2004) Comparative analysis of palm and wearable computers for part icipatory simu lations. J Comput Assist Lear 20:347-359
Kujala S (2003) User involvement : a review of the benefits and chal lenges. Behav lnf Technol 22 : 1 - 1 6
Landry M . Banvi l le C , Oral M ( 1 996) Model legit imation i n operational research. Eur J Oper Res 92:443-457
10 Part icipatory Approaches 233
Le Bars M, Le Grusse P. Al laya M, Attonaty J-M. Mahjoubi R (2004) N ECC: U n jeu de simulation pour ! ' aide a la decision collective; Application a une region medi terraneenne virtuelle. In: Pro jet IN CO-W ADEMED, Seminaire Modernisation de I' Agriculture l rriguee. Rabat
Loucks DP, K indler J, Fedra K ( 1 985) Interactive water resources model ing and model use: an overview. Water Resour Res 2 1 :95- 1 02
Maier FH, Gross I er A ( 2000) What are we talking about? A taxonomy of computer simulations to support learn ing. Syst Dynam Rev 1 6: 1 35- 1 48
Manson SM ( 2002) Val idation and verification of mult i -agent systems. In : Janssen M (ed) Complexity and ecosystem management : the theory and practice of mult i-agent approaches. Edward Elgar, Northampton, pp 63-74
Marengo L, Pasquali C ( 2003) How to construct and share a meaning for social interactions? In Conventions et Institutions: Approfondissements theoriques et Contributions au Debat Politique. Paris
Martin L et al (2007) M icroworld gaming of a local agricultural production chain in Poland. Simul Gaming 38(2) :2 1 1 -232
Mayer IS. van Bueren EM, Bots PWG. van der Voon HG, Seijdel RR ( 2005 ) Col laborative decision-making for sustainable urban renewal projects: a simulat ion-gaming approach. Environ Plan B Plan Des 32:403--423
McKinnon J (2005) Mobile interactive GIS : bringing indigenous knowledge and scientific information together; a narrative account. In : Neef A (ed) Participatory approaches for sustainable land use in Southeast Asia. White Lotus, Bangkok, pp 2 1 7-23 1
Mennet L ( 1 993) Une methode de prospect ive: Les exercices de simulation de pol it iques. Nature Sciences Societes I :34--46
Miettinen R, V irkkunen J (2005) Epistemic objects, artefacts and organizational change. Organization 1 2 :437--456
Mol in E (2005 ) A causal analysis of hydrogen acceptance. Transport Res Rec 1 94 1 : 1 1 5- 1 2 1 Moss S . Downing T, Rouchier J (2000) Demonstrating the role o f stakeholder part icipation: an
agent based social simulation model of water demand policy and response (CPM report. 00-76). Centre for Policy Model l ing. Manchester Metropolitan University, Manchester. http://cfpm.org/cpmrep76.html
Mostert E ( 2006) Part icipation for sustainable water management. In: Giupponi C, Jakeman AJ, Karssenberg D. Hare MP (eds) Sustainable management of water resources. Edward Elgar, Northampton. pp 1 53- 1 76
Nancarrow B ( 2005 ) When the model ler meets the social scientist or vice-versa. In: Zerger A, Argent RM (eds) MODS I M 2005 international congress on model l ing and simulation. Model l ing and S imulation Society of Australia and New Zealand, Melbourne, pp 38--44
Ostrom E. Gardner R, Walker J ( 1 994) Rules. games and common-pool resources. The University of Michigan Press, Ann Arbor
Pahi-Wostl C. Hare M ( 2004) Processes of social learning in integrated resources management. J Community Appl Soc Psycho! 14: 1 93-206
Pat em an C ( 1 990) Part icipation and democratic theory. Cambridge U niversity Press, Cambridge Perez P, Dray A. Le Page C, D' Aquino P, White I (2004) Lagoon, agents and kava: a companion
model l ing experience in the Pacific. I n : van Dijkum C. Blasius J, Durand C (eds) Recent developments and applications in social research methodology: Proceedings of RC33 sixth international conference on social science methodology, Amsterdam Barbara Budrich Publishers, Opladen, p 282
Ramanath AM, Gi lbert N ( 2004) The design of part icipatory agent based simulations. J Art if Soc Soc Simul 7(4). http://jasss.soc.surrey.ac.uk/7/4/ l .html
Reitsma R, Zigurs I , Lewis C, W i lson V . Sloane A ( 1 996) Experiment with simulation models in water-resources negotiations. J Water Resour Plan Manag 1 22 :64-70
Richard A, Barreteau 0 ( 2006) Concert 'eau: un outil de sociologic experimentale pour I 'etude de dispositifs de gestion locale et concertee de l 'eau. I n : Proceedings of 2" Congres de ! ' Association Fran<;:aise de Sociologie, Bordeaux. 5-8 Sept 2006
234 0. Barreteau et a l .
R ichard A, Trometter M ( 200 I ) Les caracteristiques d ' u ne decision sequentielle: Effet irreversibi l ite et endogeneisation de I 'environnement. Revue Economique 52:739-752
Rittel HW J, Webber MM ( 1 973) Di lemmas in a general theory of planning. Policy Sci 4: 1 55- 1 69 Robinson J B ( 1 99 1 ) Model l ing the interactions between human and natural systems. lnt Soc Sci J
1 30:629-647 Rocha EM ( 1 997) A ladder of empowennent. J Plan Edu Res 1 7 : 3 1 -44 Rouchier J ( 2003 ) Re-implementation of a mult i-agent model aimed at sustaining experimental
econom ic re earch: the case of simulations with emerging speculation. J Artif Soc Soc S imul 6( 4 ) . http://jasss.soc.surrey .ac .uk/6/4/7 .html
Rouwette EAJA. Vennix JAM. van Mul lekorn T ( 2002 ) Group model bui lding effect iveness: a review of assessment studies. Syst Dyn Rev 1 8 :5-45
Rowe G. Frewer LJ ( 2004) Evaluating public-participation exercises: a research agenda. Sci Technol H um Val 29(4 ): 5 1 2-556
Ryan T (2000) The role of simulation gaming in policy making. Syst Res Behav Sci 1 7 :359-364 Schel l ing TC ( 1 96 1 ) Experimental games and bargaining theory. World Polit 1 4:47-68 Schuler D, Namioka A (eds) ( 1 993) Part icipatory design: principles and practices. Lawrence
Erlbaum, Hi l lsdale Shakun ME ( 1 996) Mode l ing and support ing task-oriented group processes: purposeful complex
adaptive systems and evolut ionary systems design. Group Decis Negot 5 :305-3 1 7 Simon H ( 1 977) The new science of management decision. Prent ice-Hall , Englewood Cl iffs Star SL, Griesemer J R ( 1 989) Institut ional ecology, ' translat ions' and boundary objects: amateurs
and professionals in Berleley's museum of vertebrate zoology, 1 907-39. Soc Stud Sci 1 9 :387-420
Stennan J D ( 1 992) Teaching takes off - flight simulators for management education. OR/MS Today. pp 40-44
Ticehurst J. R issik D. Letcher RA. Newham L TH, Jakeman AJ ( 2005 ) Development of decision support tools to assess the sustainabi l i ty of coastal lakes. In: Zerger A, Argent RM (eds) MODSI M 2005 international congress on model l ing and simulation, Model l ing and Simulat ion Society of A ustralia and New Zealand, Melbourne. pp 24 1 4-2420. http://www.mssanz. org.au/mods i m05/papers/t icehu rst . pd f
Toth FL ( 1 988) Policy exercises: objectives and design elements. Simul Games 1 9:235-255 van Asselt M et al ( 200 I ) Bui lding blocks for participation in i ntegrated assessment: a review of
part icipatory methods. ICIS , Maastricht van Daalen CE. Thissen W AH, Berk MM ( 1 998 ) The Del ft process: experiences with a dialogue
between policy makers and global modellers. In: A lcamo J, Leemans R. Kreileman E (eds) Global change scenarios of the 2 1 st century: results from the I MAGE 2 . 1 model. Elsevier. London, pp 267-285
Van den Belt M ( 2004 ) Mediated model ing: a system dynamics approach to environmental consensus bui ld ing. I sland Press. Washington, DC
Yenni x J A M ( 1 996) Group model build ing, fac i l i tating team learning using system dynamics. Wi ley, Chichester
Webler T, Kastenholz H . Renn 0 ( 1 995 ) Public part icipation in impact assessment : a social learning perspective. Environ Impact Assess Rev 1 5 :443-463
W hitworth B. Gal lupe B. McQueen R ( 2000) A cognit ive three-process model of computer mediated group interaction. Gr Decis Negot 9:43 1 -456
Wood J. Si lver D ( 1 995) Joint application development. W i ley, New York
