building context in everyday life

26
JULIETTE ROUCHIER, MARTIN O’CONNOR and MÉLANIE REQUIER-DESJARDINS BUILDING CONTEXT IN EVERYDAY LIFE ABSTRACT. Social context is generally thought to influence how humans act. Here we argue that humans rarely accept the context as it is given, but rather undertake conscious actions to make it favourable. The example chosen is from northern Cameroon, where nomad herdsmen induce the sedentary farmers to trust them, by different means: creation of interpersonal links, exhibition of “good behaviours” by respecting certain norms. Trust is considered as an element of the context, necessary for them to perform acts that present a certain risk. An attempt was made to translate one of the traditional behaviours into a model, implemented in an artificial society: autonomous agents would make gifts in order to create an image of themselves in a group. In the simulations, a form of reputation stratifi- cation appears in time. One notes that an agent can build the image it wants only if the part of the population that wants to be involved in the same dynamic is big enough. As a conclusion, we suggest that although individuals try to create consciously a context for their living, this is not just an individual choice, but needs a common background of rules, a context enabling context creation. KEY WORDS: gift, multi-agent system, nomads, social simulation, trust INTRODUCTION In this paper we describe the building of trust by individuals in everyday life as a conscious creation of a favourable context. Through this analysis, we show that humans usually know, in a given social environment, how to exhibit behaviours that create the context they need to perform actions that interest them. However, it is possible to see how this manipulation of context is possible only if the majority of the population accepts certain rules. We mean here that the creation of the individual context will depend on the institutional context. The paper presents the findings of PhD research project, as an attempt to combine two very different approaches both relating to Foundations of Science 7: 367–392, 2002. © 2002 Kluwer Academic Publishers. Printed in the Netherlands.

Upload: juliette-rouchier

Post on 03-Aug-2016

216 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: Building Context in Everyday Life

JULIETTE ROUCHIER, MARTIN O’CONNOR andMÉLANIE REQUIER-DESJARDINS

BUILDING CONTEXT IN EVERYDAY LIFE

ABSTRACT. Social context is generally thought to influence how humans act.Here we argue that humans rarely accept the context as it is given, but ratherundertake conscious actions to make it favourable. The example chosen is fromnorthern Cameroon, where nomad herdsmen induce the sedentary farmers to trustthem, by different means: creation of interpersonal links, exhibition of “goodbehaviours” by respecting certain norms. Trust is considered as an element of thecontext, necessary for them to perform acts that present a certain risk. An attemptwas made to translate one of the traditional behaviours into a model, implementedin an artificial society: autonomous agents would make gifts in order to create animage of themselves in a group. In the simulations, a form of reputation stratifi-cation appears in time. One notes that an agent can build the image it wants onlyif the part of the population that wants to be involved in the same dynamic isbig enough. As a conclusion, we suggest that although individuals try to createconsciously a context for their living, this is not just an individual choice, butneeds a common background of rules, a context enabling context creation.

KEY WORDS: gift, multi-agent system, nomads, social simulation, trust

INTRODUCTION

In this paper we describe the building of trust by individualsin everyday life as a conscious creation of a favourable context.Through this analysis, we show that humans usually know, in agiven social environment, how to exhibit behaviours that create thecontext they need to perform actions that interest them. However, itis possible to see how this manipulation of context is possible onlyif the majority of the population accepts certain rules. We meanhere that the creation of the individual context will depend on theinstitutional context.

The paper presents the findings of PhD research project, as anattempt to combine two very different approaches both relating to

Foundations of Science 7: 367–392, 2002.© 2002 Kluwer Academic Publishers. Printed in the Netherlands.

Page 2: Building Context in Everyday Life

368 JULIETTE ROUCHIER ET AL.

the study of interactions. The first methodology to capture data is afield study that can be classified as ethno-economical, and the otheris the building and implementation of a computer model that wasused to perform simulations. The work was undertaken in two parts:first a field study was carried out, that inspired some hypotheseson the way society functions. Then we made a substantial liter-ature review (mainly ethnological and sociological) to enrich thequestions and a new formulation of the hypotheses on interactionmechanisms was tested by simulation of interactions in an artificialsociety. The aim was then to check if we were able to reproducedynamics that we describe.

A classical approach to interaction in human society is the onechosen by game theorists. When describing interactions, they insiston the fact that humans act with one another so that each one canachieve its interest. In this approach, the agents do not necessarilycare about the existence of the group and a relationship emergesas a result of the communications performed by the agents, withoutbeing an aim in itself. Another way of understanding the relationshipthat is created is to show how the individuals consciously build upthe context of the interaction. In particular, most of the behavioursthat take place in any meeting are directly aimed at making that verymeeting possible (Goffman, 1961). Each one is concerned by hisown appearance as it is perceived by the other, and does his best tosecure the link: roles can be recognised, reproduced or changed ina conscious way. A very important element that helps in strength-ening the relations, and makes them persist in the future, is thepresence of trust between people (Luhmann, 1979; Dasgupta, 1988).This is particularly true if some merchant exchanges (trade) are tobe performed: each actor must believe that he can trust the otherand/or wants the other to trust him. There exists then several waysto create a trusting context among the agents: some mechanismsare used in a first meeting (and consist mostly of the exhibitionof guarantees recognised by a social group) (Livet and Reynaud,1995) or are developed through time (when familiarity grows inrepetitive meeting (Marsh, 1994), or when an individual builds uphis reputation in a group).

The use of multi-agent models is now beginning to be quitecommon in social sciences: it is the translation of some social

Page 3: Building Context in Everyday Life

BUILDING CONTEXT IN EVERYDAY LIFE 369

dynamics observed into a multi-agent system, in order to performsimulations with a computer (Conte and Gilbert, 1995; Epstein andAxtell, 1996). Multi-agent models are very interesting models touse for social scientists since they allow very precise descriptionsof interactions and of their repetition. By having simple agentsrepeat the interactions they can perform, it is usually possible towitness the emergence of some social patterns in groups, regular-ities or hierarchies (Doran, 1994; Rouchier et al., 1998). The resultsshould then give some insight on the issue addressed and in partic-ular enrich the perception of the interactions in the group, as it hasalready been shown in various other “mixed” works of the same kind(Barreteau and Bousquet, 1999).

First there is a description of the results of a field study that wascarried out in Northern Cameroon among transhumant herdsmen,where the issue of building an appropriate context for everydaylife revealed itself to be very important. Indeed, trust is necessaryto have access to water as well as to make merchant exchanges,and making gifts turned out to be one of the most important waysof inducing that trust in the local population. Indeed, if trust canindeed be established in this way, it is essentially because the respectfor some form of normative behaviour does create a reputationfor the individual who performs it. The model that was built isdescribed: it deals with the emergence of an image in a group,thanks to the occurrence of the very single act of giving. Simulationswere performed in the artificial society based on the model. Theyled to the conclusion that it is indeed possible to reproduce quitesimply the process of building some reputation for an individual.But it was also found that one element of great importance wasthe participation of a minimum number of agents to that dynamicsof consciously building reputation. The overall conclusion of thatwork is that although the creation of context represents an importantpart of the actions that the individuals take in everyday life, no onecan ever consider it as the action of one individual, but it has to beunderstood as an interacting process. Unless a significant part of thegroup is ready to use the same codes of communication or wishesto produce the same kind of context a single individual will not beable to succeed.

Page 4: Building Context in Everyday Life

370 JULIETTE ROUCHIER ET AL.

GIFTS: NORMAL BEHAVIOUR, RATIONAL INTEREST ORWHAT . . .?

Social Environment

The field study took place for three months in the Extreme North ofCameroon, mainly in the Kaele region, in Maroua and in the Waza-Logone plain, and investigated the issue of access to the naturalresource for nomad herdsmen. To achieve the study, interviews wereheld with transhumant herdsmen and farmers. The farmers inter-viewed were of two “categories”: some were in villages where thenomadic population come very regularly and are well known. Someothers were in villages in which very few nomads are usually seen,mostly because they usually stay in the bush and do not mix withlocal people, of rather different ethnic groups.

Meeting different people gave the opportunity to hear very diver-gent opinions about the way the bush should be used, and who mayor may not use the resource. Although the perceptions were veryextreme, it is possible to identify some elements in common aboutthe bush, and the main one is the risk that everybody perceives ofbeing far from villages (which has been true all along history inseveral areas where people do transhume (Spencer, 1973)). At themoment, there are a lot of bandits who circulate quite freely in thearea and who got their weapons during the Chad war a few yearsago. They usually assault people who are isolated in the bush, butcan also join in a big troop so that they can attack a whole camp ora taxi-brousse (bus) that travels between cities.

The population that mainly practices transhumance belongs to anethnic group called Peuls or Fulbe. That population is widely spreadin the area, and everyone has relatives living as sedentary farmers invillages (Dupire, 1962). If the herdsmen have to move during theyear, it is because they possess large herds of cows. The huge vari-ability of the weather and the resource makes it impossible to keepthe animals in only one area for the whole year. The herdsmen andtheir employees must find places to keep the herd, where it can haveaccess to water and grass. Hence, for a least a few months duringthe year, the owner leaves his animals in the bush near villages towhich he does not belong. Different people, although having verydifferent activities (like hunting, gathering fruits or growing crops)

Page 5: Building Context in Everyday Life

BUILDING CONTEXT IN EVERYDAY LIFE 371

use resources that are in a shared area. This is why there must exista common agreement with the authority of the villages (usually atraditional chief) in order to decide if their presence is accepted (andit is traditionally the case that the chief always accepts). The mainrisks appear during the cropping period, when the cows can wanderand eat the cereals in the fields. But even if the herdsmen have toavoid conflicts that could appear, they also have to stay in walkingdistance from villages so that they can go to the market, either tosell animals or milk or to buy some food or goods.

On Some Habits and Their Justification

What is usually expected from the people that stay near a villagefor a few weeks is that they go and see the chief, possibly makinghim a gift, and to tell him that they have just arrived and ask himfor his support. This is the traditional way of doing it, and is usuallypresented as the fundamental act that creates the relation betweenthe herdsmen and the local communities. This behaviour is stillquite important in the region, although it is not necessarily easy tounderstand in the contemporary context. The official power that ispresently in place is no longer supposed to be the traditional one andthe power of the chief is now unofficial and seriously weakened, andthe traditional structures are less reliable for the people. During thefield study one question asked of the people involved in these inter-actions was their reason for following this tradition. Two answerswere generally given to the survival of that practice:

The first one referred to the norms of behaviour just described.Some people explained that they had to follow the rules thatwere traditionally defined, and that they always would. They didnot indicate at first that it could occur to them not to follow it.What seemed to be quite clear, though, after a few questions, isthat all people choose to what extent they would respect it, bydeciding personally in which precise circumstances the rule has tobe respected. For example, people can choose to go the chief onlyif they stay for more than a few weeks on his land; they choosehow much to give: it is usually a symbolic gift (sugar and teabeing the minimum), except for rich people who tend to reaffirmtheir importance by offering at least the value of a cow. There areanyway exceptions to these loose rules: some people never go and

Page 6: Building Context in Everyday Life

372 JULIETTE ROUCHIER ET AL.

tell the chief that they are on their land (this is not true for Peulsbut rather for herdsmen from other ethnic groups); some very richpeople, when they don’t like a chief, make very small gift, which isinterpreted as a lack of respect for all the people who know about it.

The second answer is of a quite different kind, and refers to amuch more self-interested perspective. Some herdsmen said goingand seeing the chief is mainly motivated by the wish to make an allyof that man. What is indeed suggested by that gift is the request tobe helped in case there is any problem, and that another gift couldcertainly be made in that circumstance: it is an attempt of corruptionthat people use quite consciously. Although most herdsmen justifytheir act by the insurance that they thus get, they can also give a lotof counter-example showing that the protection induced is purelytheoretical. The support of the chief can be useful in two mainscases: if the herdsmen are attacked, the chief would feel obligedto send his men to catch the bandit and punish him; or if therewere some disagreement between the nomads and some people fromthe local population over the use of the resource (most cases beingwhen a cow goes on a field). But it was also reported many casesin which the protection was not guaranteed (the chief having notenough power to protect or deciding to help farmers from the villagerather than the herdsmen whatever had happened), and these eventswere indicated as becoming more and more frequent.

Actually, what happens most of the time is that people try neverto go and see the chief when there is a problem, although theyhave “bought” his protection. What they rather do is to sort out theproblem with local people, who sometimes fear their chief enoughto like that solution better. For us, both arguments presented by thepeople give only part of an answer to the question of why theyconsciously choose to respect an old norm of behaviour. Actuallywhat seemed to be much more important for them than their relationto the chief is the opportunity to have amicable negotiations. Andthis kind of negotiations can actually occur only if at least a part ofthe population has a good image of the herdsman, which means: ifthe context is positive towards that person.

Page 7: Building Context in Everyday Life

BUILDING CONTEXT IN EVERYDAY LIFE 373

Building the Context for Amicable Negotiation

The different points of view of local farmers regarding the herdsmenwere good indicators to understand how the behaviours of theherdsmen, however mundane, had an important effect on the localpopulation. Some acts seemed to have no direct effect, but they wereactually useful (and maybe more or less consciously intended) toestablish and reproduce the context in which the social interactionswere to take place.

Where trust is needed is for several kinds of actions that are verynecessary to everyday people’s life:

• When animals are sold (because a farmer wants to make surethat the animal he buys is not ill and a herdsman wants to besure he will be paid even if he allows a delay,

• When a farmer pays a nomad to keep his animals (which is aquite common practice): this needs a huge amount of trust orthe belief that one can find the person again,

• In cases of conflict when it seems suitable to get to an amicableagreement to solve the problem, it is necessary that the farmershave a good image of the herdsman to make the negotiationpossible.

Different elements seem to be necessary for a farmer to undertakeone of the three “risky” interactions with a herdsman. The mostusual reason is the existence, for a certain period, of a previouspersonal link. This is not necessarily a direct link, since in thatculture the relations are very transitive: the person can be knownvery well by members of the family or by friends. The other wayof building an image of a nomad, that could generate trust, is bywatching his behaviour over time or hearing what people say abouthim. If a herdsman comes regularly he usually has in the village areputation that is based on his past behaviour that has been observedby the whole population. To have a good image, having personallinks with members of the local community that can indicate hissocial status and reliability. But the farmers are also quite sensitive toostentatious behaviours of a herdsman: if he shows that he respectssome basic social rules, he is more likely to be trusted, even ifnobody knows him yet.

Page 8: Building Context in Everyday Life

374 JULIETTE ROUCHIER ET AL.

One of the most awaited actions is the visit to the chief and thegift that can be made on that occasion. Even if the farmers do notsee the herdsmen come to the village, he generally hears the reportsof witnesses who make comments about the event. Any nomad whowould stay in the bush without coming to say where he installed hiscamp is necessarily thought to be a thief by the local population.Another attitude can be quite important: if some of the memberscome regularly to the village, on market days for example, the wholecamp is very likely to be accepted. Witnessing regular appearancesof herdsmen and knowing that they came to see the chief showsthat they are willing to be known and to participate in the socialdynamics of the group. These elements have been observed by quitea lot of ethnographers and reported as tacit understandings betweenpopulation that rely on the emission of expected signals of that sort(Guichard, 1998; De Bruijn and Van Dijk, 1997).

The gift was difficult to explain just by the fact that it is a norm,since it was very far from being automatically respected. It wasnot simply explained in terms of self-interest, since it was ofteninsufficient as a mean of corruption and quite rarely used as. Butthese two approaches seemed to be quite well extended by the ideaof an implicit creation of context (Nicolas, 1986). The herdsmenjust know that it is better for them to show as overtly as possibletheir respect of the rule to the community they will have to dealwith. Here we consider the respect of these rules of behaviour, as amore or less conscious building of reputation. The reputation thatthe herdsmen have in a village is here taken as one of the mainelements of the context in which they have to live. Their consciousbuilding of their reputation is the preservation of a positive contextin the environment they have to live in.

The process of creating reputation thanks to public acts seemedto be important enough to be studied in itself, and this is why wewanted to build a model to represent it. The aim of the model wasto consider the building of a reputation that would depend only onwhat agents could observe of each other, without using interpersonalrelation as evidences.1 Here, we didn’t stick to the reality of thefield, which complexity was too high to be represented accuratelyin a model, but we decided to focus on one public act, the gift.Studying that type of interaction we decided to use mainly a liter-

Page 9: Building Context in Everyday Life

BUILDING CONTEXT IN EVERYDAY LIFE 375

ature review (and not so much the field data) to represent the giftas an ostentatious act. By the simulation, we then wanted to test theprocess that we thought we had understood. If it is really possible foragents to make themselves different from others (in reputation) justby performing actions, then maybe the described process would benot too far from reality. In the next section the model is described,as well as the choices of simulation and of observation that weremade and the conclusions the results lead us to (also see (Rouchier,2000)).

AN ARTIFICIAL SOCIETY DESCRIBING A MODEL OFREPUTATION

General Issues

The aim in building the model was to describe a society of inter-acting agents where the act of each one could be observed by theothers, and would lead all of them to have a representation of theothers. In artificial societies, it is not recommended to study toomany processes at the same time, since it is easy to get results thatare more complex than the studied system, and where the impor-tance of parameter and of arbitrary choices is not manageable. Thatis why we decided that reputation would be based on the sameelements for the whole population. The results of the field studyindicated the importance of the act of giving, which has severallinked meaning and dynamics. This is why we choose that act asthe representation of a deeply socially situated act. An advantageof that choice was also that the ethnological literature is quite wideconcerning the “societies of gifts”, since they display very importantsocial dynamics (Malinowski, 1922; Nicolas, 1986; Bateson, 1958).

In all societies, giving is an important act, but in some it is themost important exchange performed. In these societies, it reallygives a direct meaning to the circulation of goods by generatingstatus in a group, whereas the merchant exchange has less directinfluence on the definition of the individual role (Godelier, 1996).There are then a few elements that need to be stressed to see how anindividual is supposed to behave in that kind of society:

• There exist several reasons to do a gift: it can be made for re-distribution purposes, to help someone who needs goods at one

Page 10: Building Context in Everyday Life

376 JULIETTE ROUCHIER ET AL.

moment; it can be made to create a link since it creates thefeeling of being dependent on someone; it can be made to showthat one has no interest in goods and can get rid of it easily,which expresses a great power.

• To make gifts regularly is necessary to be recognised as part ofthe social game since it is the main way to share the goods withothers or to re-affirm a prestigious position (which is called anostentatious behaviour). There even exist real fights betweenchiefs when it comes to giving: some express their power byforcing their opponent to receive gifts; they hope the otherwon’t be able to give back and will thus feel humiliated (Boas,1966).

• There are protocols defining the way the gifts have to beperformed to become social acts: to whom it must/may bemade, on what occasions, how much should be given. Usuallyone shares with people who are less important and rich and oneprovokes someone who is at least as prestigious.

Most of these elements were indeed observable in NorthCameroon to define the gifts that the herdsmen were supposed todo. There is one element that we didn’t want to represent here: theidea of a gift made to create a debt between people and thus toreaffirm a link.2 It is only the meanings that could help to createthe image of an individual in the group that we wanted to use. Herewe represented giving as the only act that the agents are able todo, and feel like they have to. There are two gifts: one is just theparticipation to the normal exchange dynamics (called the “sharinggift”), and the other is an attempt to get prestige in the group (thislatter, “prestige gift”, is harder to perform).

As is usual for multi-agent models, the virtual universe isdescribed as an environment with different agents that interact witheach other (Ferber, 1999; Conte and Gilbert, 1995). Each one hasan explicit representation of its environment and of itself that makesit able to act: it takes information in the environment and computesit to know what it should do in given circumstances. A simulationis a succession of time-steps in the artificial world, beginning withan initial setting. During each time-step, the agents can undertakecertain actions and this imply an evolution of the world. Agentsinterpret the change of environment and usually perform another

Page 11: Building Context in Everyday Life

BUILDING CONTEXT IN EVERYDAY LIFE 377

act at the next time step, since they are in a different environment tochoose. Sometimes the agents keep the same methods for choosing,but it is also quite common that the agents should have their choiceprocesses evolve due to the environment changes. We performedsimulations so that to find the minimum element that could producea society with a reputation due certain norms of behaviour. Wewanted then study how the heterogeneity of the agents in terms ofinterests and representation of themselves could have an influence.Eventually we wanted to see if the results of the model could beused to reconsider some elements of the field study.

Agents and Group: Characteristics and Interactions

There are 50 agents that are gathered in a group. This group isresponsible for the building of the common image, as well as thedistribution of the gifts among the agents. The choices of behaviourfor each agent is defined by two parameters:

− Its motivation (either sharing with others or trying to getprestige),

− Its self-esteem (which helps it to decide if it is able or not toperform the gift it wants).

A time-step goes as follows:All agents choose the gift they want to make according to

their self-esteem, their motivation and a randomised element (seeFigure 3):3

• Test for motivation: if (Random∗motivation for prestige –Random∗motivation for sharing) >0, the choice is prestige.4

• Test for esteem: if (Random∧2∗10) > self-esteem, it cannotperform the gift it wants to.5

• Test for money: if money > (3∗minimum consumption) it canperform a prestige gift; if money < (2∗minimum consump-tion), it must work to be able to do a sharing gift.

When all gifts have been chosen, the group designates randomlyto which agent each gift is going to be given, knowing the kind ofgift and the classification of the agents:

• If an agent of rank R makes a sharing gift, it will go randomly6

to any agent of rank <= (R + 2).

Page 12: Building Context in Everyday Life

378 JULIETTE ROUCHIER ET AL.

Figure 1. A gift is an object that circulates among the agents.

Figure 2. Characteristics of the agents and actions they perform.

Page 13: Building Context in Everyday Life

BUILDING CONTEXT IN EVERYDAY LIFE 379

Figure 3. The whole process of choice for an agent: several tests and constraints.

• If an agent of rank R makes a prestige gift, it will go randomlyto any agent of rank >= (R − 2).

Then the gifts are distributed: the money goes from one agentto the other and knowing what each agent gave and received, thegroup calculates the new reputation, and thus the ranks (which isjust a classification of the agents in terms of prestige reputation: theagent with rank 1 has the highest reputation prestige in the group).The reputation is the sum over the last 25 years of the number ofcredits for each category (Figure 4).

Simulations and Results

Several kinds of simulation were performed, whose aim was to testthe possibility of creating reputations for the agents. What we wantto witness is a differentiation between the agents in terms of reputa-tion, and possibly the conditions for stability of the ranks (or a never-ending instability).

The system has been tested in simulations where the agents hada stable self-esteem (here a representation of their ability to act) andof their motivation for prestige. The agents were not necessarilyhomogenous, and they sometimes were separated in groups with

Page 14: Building Context in Everyday Life

380 JULIETTE ROUCHIER ET AL.

Figure 4. The calculus of reputation at the end of each time step.

high and low self-esteem, or groups with high and low motivationfor prestige. What the simulations revealed is that the ability to builda specific image of itself for an agent depends not only on its moti-vation and self-esteem, but also on the behaviour of the others, in avery strong way. And actually it is possible to identify some patternsof societies that can emerge, depending on the number of agents thathave the same self-esteem or the same motivation for prestige.

In the societies produced it is interesting to capture the patternsthat can be observed and more precisely the regularities in thereputation of some agents and the possible apparition of sub-groups that would be more or less stable. This is why the maincharacteristics that were observed were:

− Individually: the evolution of rank over time for each agent.− Globally: a picture of the society where each agent is repre-

sented by its sharing reputation as a function of its prestigereputation, at one time-step.

− Globally: the distribution of gifts in the group at one time-step.

Page 15: Building Context in Everyday Life

BUILDING CONTEXT IN EVERYDAY LIFE 381

Figure 5. The first figure shows the structure of the population at one time-step:the sharing reputation is given as a function of the prestige reputation for eachagent: one sees an elite with a high prestige reputation that is differentiated.

Homogenous groupWe first made simulations where all the agents had the same moti-vation and same self-esteem where we tested the values of eachto see what would happen. Actually it was possible to see somevery different shape of society depending on the value chosen. It ispossible to identify groups that are more or less stable (Figures 5and 6), or to have no differentiation between the agents, dependingon the values chosen (Figure 7). The results given here are quicklyexplained and can be found with more details in Rouchier (2000).

Actually the results when the agents all have the same self-esteemand motivation can be summarized in:

Motivation for prestige<= 5

6 <= Motivation <= 9 Motivation for prestige= 10

Self-esteem>= 6

No differentiation, highvariation of rank for allagents, with high repu-tation for prestige.

An elite appears, whichis quite stable.

The elite is not verydistinct and the agentshave quick changingranks.

Self-esteem<= 6

No differentiation, lowreputation for all.

An elite appears and itchanges often.

An elite appears, thatchanges often.

Page 16: Building Context in Everyday Life

382 JULIETTE ROUCHIER ET AL.

Figure 6. The second figure shows the rank of two agents during 1000 time-steps,showing that an agent can stay in the elite (rank with low number) for a quite longperiod.

Figure 7. No agent or group that differentiate itself from the others, in term ofprestige reputation (again: here one can see the sharing reputation as a functionof prestige reputation).

Page 17: Building Context in Everyday Life

BUILDING CONTEXT IN EVERYDAY LIFE 383

The main conclusion that has to be stressed about the simulationperformed is that if we want to picture an ability to differentiatefor the agents, we must put a minimal motivation for prestige and aminimal self-esteem. But at the same time, if too many agents havethe wish to get prestige and feel that they are able to do so the elitethat appears can never be stable.

Heterogeneous groupThen we made some simulation with two different subgroups for theagents:

• Either with the same motivation and two self-esteems (4 or8, depending on the group). The motivation for prestige wasvalued to 8 so that to be in conditions were a differentiationwould appear in general.

• Or with the same self-esteem and two motivations (10 or 4,depending on the group). The self-esteem was valued to 8 sothat the agents make a lot of gifts.

When the self-esteems are different, it means that some agentsfeel more able to perform the gift they choose than others. If onlyone agent has a high self-esteem, the system exhibits a society wherealmost no gift happen and without any elite (even the agent withhigh self-esteem cannot stay with a high reputation). If less than 3agents have a high self-esteem, they constitute an elite which is verystable, and in which agents with a low self-esteem are sometimesincluded (Figure 8). If the agents with high self-esteem are morenumerous but less than 10, the size of the elite is still of 3 or 4 andit is still very stable, but no agent with low self-esteem can get init anymore. When more than 11 agents have a high self-esteem, thesituation is similar to a simulation with a homogenous populationmotivated for prestige: there is an elite in which all agents have highself-esteem and it is not very stable, since all agents go out of theelite in less than 200 time steps. In most of these simulations, thesituation is pretty predictable: the agents with high self-esteem arethe one who tend to be in the elite because they are the only onethat can regularly do the prestige gift that they want to do. But onecan notice that there are agents with low self-esteem in the elite atall time-step. The reason is that the dynamic is very social in thatsystem and it is possible to benefit from the others’ acts.

Page 18: Building Context in Everyday Life

384 JULIETTE ROUCHIER ET AL.

Figure 8. A simulation where four agents have a high self-esteem. These agentssucceed in being in the elite, but they drag with them some agents with a lowself-esteem, that receive lots of gifts of prestige and thus have a prestigious rank(there are two of them at that point).

Even more surprising is the situation where agents have differentmotivation for prestige. The motivation is what expresses the valuethat the agents give to the action of making prestige gifts (withhigher value and more difficult to do than the sharing gifts). If only afew agents (less than four) have a high motivation for prestige, thereis not possibility for them to constitute a stable elite. They succeed inhaving a higher reputation at some moment, but it is not so far fromthe one of other agents, and it is a very unstable situation (Figure9). If there are between 4 and 15 agents who want the prestige, astable elite appears, as well as a second group constituted of agentswho want the prestige, but these can never get to a prestigious rank,because the others keep the places in the elite (Figure 10). If thereare more than 16 agents wanting the prestige, the situation is verysimilar to the one with all agents having the same motivation forprestige.

Page 19: Building Context in Everyday Life

BUILDING CONTEXT IN EVERYDAY LIFE 385

Figure 9. A simulation where two agents have a high motivation for prestige.One of them succeeds in having the highest prestige reputation at that time step,but the other is in the second half of the population. There is no elite clearlyappearing, just a continuum on agents being close to each other in reputation.Only two agents wanting the prestige cannot generate a differentiation for theprestige in the society.

Conclusion

We can conclude from the simulations with different groups ofchoice mechanisms that there is no direct correlation between therank of an agent (or its prestige reputation) and his motivation orself-esteem (that determine his choice of action): sometimes agentswho do not want prestige get it, and some who want it are unable tohave it during the whole simulation. Even without high self-esteemit is possible to be in the elite because the others need you in theirdynamics. If not enough agents have the motivation to have theprestige, the one would want it are not able on themselves to havethe system work: an agent on his own cannot build its reputation,even if he follows the rules, but is dependent on the wishes of theothers too. If a certain number of agents want the prestige, only fewcan get it, because the others block the access to the highest ranks.

Page 20: Building Context in Everyday Life

386 JULIETTE ROUCHIER ET AL.

Figure 10. In that simulation, eight agents want the prestige.

What can thus be concluded with this example is that a dynamicsthat involves the observation of the others and a necessary inter-action with them is dependent of the repartition of certain desiresand beliefs in the group. Especially, in all the simulations, whennot enough people are motivated by prestige, the wish to differen-tiate oneself never leads to an efficient result. Even if one follows aspecific rule, if this rule does not stand as an important value for theothers, the awaited result cannot appear.

DISCUSSION

It seems that trust is in some cases a very important element of socialrelations so that any serious interaction can take place. Here, wepointed out through an example a quite important way of buildingtrust, which is the “emission” of special signals through ostentatiousbehaviour in a society with strong norms. In the North-Cameroon,it seems that the respect of norms turns out to be a message initself, and that the herdsmen use it when they intend to build afavourable context for their interactions in a village. This conclusion

Page 21: Building Context in Everyday Life

BUILDING CONTEXT IN EVERYDAY LIFE 387

can be linked to others that were drawn after observations made insome other places (Goffman, 1961). This type of mechanisms canindeed be recognised European societies as well: some behavioursare identified as the sign that the new comer knows the rules andaccepts to play a certain social game (Gambetta, 1988; Boltanskiand Thevenot, 1987). These European behaviours are anyway quitedifferent from the one that were here identified, and among whichwe designated gift as typical.

Having identified the importance of the reputation created byostentation, we tried to reproduce that phenomenon in an artificialsociety, to see if the observation of its actions could indeed createan image of an agent for the other that could be significant. Here thesignificance would be defined by the ability to be clearly differentfrom the other or to be in a stable position (for a period that wouldbe significantly longer than the memory time-path chosen). The firstresult was that it was indeed possible to create a differentiationamong agents. A second point that was positive is that the individualhistories are not easily predictable with the rule we put in it, even ifthe logic seems very circular. Indeed, the actions of the whole grouphad to be taken into account for the reputation of each individual:the intention of an individual agent to get prestige (that we called“motivation”) was not enough for it to have it get to the position itwants. Its actions had to be sustained by the intention and actions ofother agents. In any case, it was extremely difficult to dissociate theactions of an individual from the expectations that the others in thegroup can have. This reminds us that to build a reputation, there is aneed of having a shared interest in certain values.

The observation of processes that people use to build trust isusually an approach that helps to capture the real beliefs of theindividual about the social norms (Dasgupta, 1988). For example ithelps to discover the sanctions that people expect to be applied andthose they don’t rely on (Hechter, 1990). To be able to create a socialcontext in which they can evolve, people must necessarily knowwhat is theoretically expected. But that knowledge is not enoughif, at that moment the other actors are not ready at that moment toparticipate in the same dynamics. In that case the actions fail to havea meaning. Indeed the institutional context7 evolves in societies, and

Page 22: Building Context in Everyday Life

388 JULIETTE ROUCHIER ET AL.

even in one year a rule can be not as significant just because someevents have destroyed the previous agreements.

This brings us back to the behaviours of the transhumant peoplein North-Cameroon. In the sahelian zone, up until recently, therules of behaviours were pretty stable, and identical in a large area(Monod, 1975). The traditional structure has been slowly destroyedby the new official structures that were installed with the colonisa-tion: the superposition of different powers makes it uneasy to have acoherent set of rules as there used to be (Toure, 1990). To knowwhich rules to apply in an area or which signs are going to beinteresting to the locals, herdsmen have to be well informed. Thisis why it is increasingly difficult for the herdsmen to change theirdestination (like going to Nigeria instead of Cameroon), be it for ayear or for good, if they have no friends or relatives in the new area(Requier-Desjardins, 1997).

The situation described here is one where people are used tobeing migrant. The population is constituted of people whose familytradition is related to nomadism: they have a strong knowledge ofhow to evaluate the need of communication and how to formaliseit. During interviews we had indeed noticed that herdsmen with anethnical background that was more sedentary, had a higher chanceto underestimate the importance of the rules. This also raises themore general question of migrants who do not necessarily have aculture of movement and who have no knowledge of the relevantinstitutions that can enable them to create trust. Without a hugeimplicit knowledge it is difficult to build a favourable context andbe accepted. But making explicit the expectations of the group isnot central in the everyday actions of its members. This is certainlywhy suspicion appears so often because of small misunderstandingin the interpretation of actions. Not knowing the general context ofcommunication and it subtleties makes people unable to build thelocal context that would allow them to be accepted.

Since a contemporary issue seems to be able to secure the life ofmigrants,8 one important here seems to be the stability (or at leastthe visibility) of expectations of the local group. If the institutionalcontext itself evolves very quickly and with no homogeneity, it ismuch more difficult for people to manipulate the social environmentin which they can live. If the rules are unclear and nobody pays

Page 23: Building Context in Everyday Life

BUILDING CONTEXT IN EVERYDAY LIFE 389

attention of the efforts to create trust, even with very good intentionsit will be hard to be trusted.

The use of multi-agent simulation was actually very useful inthat work. Indeed, it gave us a point of view that was mainly inter-actionist and thus helped in the exploration of the field. By testinghypothesis in simulations, it enabled us to formalise very clearly theassumptions we were doing. It was also a very good tool to trackin our system the role of each individual in the global dynamics,how each of them could be limited or helped by the actions ofothers. The notion of trust along with that tool indeed gives aninsight of a double relation that is not easy to track precisely: theimportance of the society in individual action (here, through officialnorms of behaviour), and the emergence of social structure thanksto individual actions.

ACKNOWLEDGEMENTS

Thanks to François Bousquet, Christophe Le Page and OlivierBarreteau whose example gave me enough patience to study simu-lation results; and to Jacques Weber who insisted in sending us tothe country he loves.

The model has been developed with VisualWorks 3.0, theSmalltalk-80 environment produced by ObjectShare. The researchreported here was funded by the CIRAD (French centre in agronomyand development) and by the French Ministry of Research.

NOTES

1. The study of the interpersonal link between herdsmen and farmers has led toanother model in which familiarity is represented, called JuMel (Rouchier etal., 2001).

2. The relationships and dependence that are created thanks to debts in a giftsociety have been studied in a previous model (Rouchier and Bousquet,1999).

3. The randomised element indicates that the determinism is not absolute. Wetried to stick to a intuitive point of view that would say that even someonevery motivated by prestige can sometimes do sharing gift, and someone veryself-assured could sometimes feel unable to do an act.

4. Motivation for prestige + motivation for sharing = 10.

Page 24: Building Context in Everyday Life

390 JULIETTE ROUCHIER ET AL.

5. 8 > esteem > 2.6. Here the random element emphasizes the idea that people do not create

relationship through that gift but just participate in a social dynamics.7. An institution being the “set of rules in use” (Ostrom, 1990).8. At least for those who acknowledge the fact that transhumance and migration

are phenomena that are culturally and economically important and that theyshould not (or could not) be refrained (Boutrais, 1995; Behnke and Scoones,1992).

REFERENCES

Barreteau, O. and F. Bousquet: 1999, SHADOC: A Multi-Agent Model to TackleViability of Irrigated Systems, Annals of Operations Research 94: 139–162.

Bateson, G.: 1958, Naven: A Survey of the Problems Suggested by a CompositePicture of the Culture. Stanford, California: Stanford University Press.

Behnke, R.H. and I. Scoones: 1992, Repenser l’écologie des parcours: implica-tions pour la gestion des terres de parcours en Afrique, Dossier du programmeréseaux des zones arides, 33, IIED. London, England.

Boas, F.: 1966, Kwakiutl Ethnography. Chicago: The University of Chicago Press.Boltanski, L. and L. Thévenot: 1987, De la justification: les économies de la

grandeur. Paris: Gallimard.Bousquet, F., P. D’Aquino, J. Rouchier, M. Requier-Desjardins, A. Bah, R. Canal

and C. Lepage: 1999, Rangeland Herd and Herder Mobility in Dry Intertrop-ical Zones: Multi-Agent Systems and Adaptation People and Rangeland. InD. Elridge and D. Freudenberger (eds.), Building the Future: InternationalRangeland Congress, 6, Vol. 2, 831–836.

Boutrais, J.: 1995, Hautes terres s’élevage au Cameroun, Etudes et thèses. Paris:Orstom.

Conte, R. and N. Gilbert (eds.): 1995, Artificial Societies. The Computer Simula-tion of Social Life. London: UCL Press.

Dasgupta, P.: 1988, Trust as a Commodity. In D. Gambetta (ed.), Trust. Makingand Breaking Cooperative Relations. Oxford/New York: Basil Blackwell,chapter 4, 49–72.

De Bruijn, M. and H. Van Dijk (eds.): 1997, Peuls et mandingues. Dialectiquesdes constructions identitaires. Paris: Karthala.

Doran, J. M. Palmer, N. Gilbert and P. Mellars: 1994, The EOS Project: ModellingUpper Paleolithic Social Change. In N. Gilbert and R. Conte (eds.), SimulatingSocieties. The Computer Simulation of Social Phenomena. London: UCL Press,103–125.

Dupire, M.: 1970, Organisation sociale des Peul. Etude d’ethnographiecomparée. Paris: Plon.

Epstein, J.M. and R. Axtell: 1996, Growing Artificial Societies. Social Sciencefrom the Bottom Up. The Brookings Institution, USA.

Page 25: Building Context in Everyday Life

BUILDING CONTEXT IN EVERYDAY LIFE 391

Ferber, J.: 1999, Multi-Agent Systems. An Introduction to Distributed ArtificialIntelligence. Harlow, England: Addison-Wesley.

Gambetta, D. (ed.): 1988, Trust. Making and Breaking Cooperative Relations.Oxford/New York: Basil Blackwell.

Godelier, M.: 1996, L’énigme du don. Paris: Fayard.Goffman, E.: 1961, Encounters: Two Studies in the Sociology of Interaction.

Bobb-Merril.Guichard, M.: 1998, L’étrangeté comme code de communication inter-ethnique:

Des relations entre agropasteurs Fulbe et paysans Bariba du Borgou (Nord-Bénin). In Y. Diallo and G. Schlee (ed.), Dynamique des Frontières: Del’ethnicité Peule dans de nouveaux contexts. Paris.

Hechter, M.: 1990, The Attainment of Solidarity in Intentional Communities,Rationality and Society 2(2), 142–155.

Livet, P. and B. Reynaud: 1995, La confiance indécidable et ses versions enéconomie. In Actes Du Séminaire Interdisciplinaire: “Confiance, apprentis-sage et anticipation économique”. Compiégne: Université Technologique deCompiègne, 208–222.

Luhmann, N.: 1979, Trust and Power. London: John Wiley and Sons.Marsh, S.: 1994, Formalising Trust as a Computational Concept. PhD Thesis,

Department of Computing Science and Mathematics, University of Stirling.Monod, T.: 1975, Introduction. In T. Monod (ed.), Les sociétés pastorales en

Afrique. Proceedings of the XIIIth International African seminar, Niamey,1972, International African Institute, Oxford University Press, London, Ibadan,Nairobi, 298–321.

Nicolas, G.: 1986, Don rituel et échange marchand dans une société sahélienne.Paris: Institut d’ethnologie.

Ostrom, E.: 1990, Governing the Commons. The Evolution of Institutions forCollective Action. Cambridge, GB: Cambridge University Press.

Requier-Desjardins, M.: 1997, L’accès aux pâturages, Une approche économiquede la mobilité. In Proceeding of the conference “Méga-Tchad, L’homme etl’animal dans le bassin du Lac Tchad”. Paris: Orstom.

Rouchier, J., F. Bousquet, M. Requier-Desjardins and M. Antona: 2001, A Multi-Agent Model for Describing Transhumance in North Cameroon: Comparisonof Different Rationality to Develop a Routine, Journal of Economic Dynamicsand Control 25: 527–559 (Springer).

Rouchier, J.: 2000, La confiance à travers l’échange. PhD Thesis, UniversitéOrléans, France.

Rouchier, J. and F. Bousquet: 1999, Non-Merchant Economy and Multi-AgentSystem: An Analysis of Structuring Exchanges. In J. Sichman, R. Conte and N.Gilbert (eds.), Multi-Agent Systems and Agent-Based Simulation, LNAI series,Vol. 1534. Springer, Berlin: Springer-Verlag, 111–123.

Rouchier, J., O. Barreteau, F. Bousquet and H. Proton: 1998, Evolution andCoevolution of Individuals and Groups. In Yves Demazeau (ed.), Proceedingsof the Third International Conference on Multi-Agent Systems. Los Alamitos,USA: IEEE, 254–260.

Page 26: Building Context in Everyday Life

392 JULIETTE ROUCHIER ET AL.

Spencer, P.: 1973, Nomads in Alliance. Symbiosis and Growth Among the Rendilleand Samburu of Kenya. Oxford University Press.

Toure, O.: 1990, Ngaynaaka Majji: la perte des pratiques pastorales dans leFerlo (Nord Sénégal), Programme réseaux des zones arides, 22. London: IIED– International Institute for Environment and Development.

Centre for Policy Modelling Juliette RouchierMMU, Manchester

C3ED, Université de Saint-Quentin-en-Yvelines Martin O’ConnorFrance

CIRAD-TERA, Montpellier Mélanie Requier-DesjardinsFrance and C3ED