behaviour in commons dilemmas: homo economicus and homo ...€¦ · studying how micro-level...

Ecological Economics 35 (2000) 357–379

SPECIAL ISSUE

THE HUMAN ACTOR IN ECOLOGICAL-ECONOMIC MODELS

Behaviour in commons dilemmas: Homo economicus andHomo psychologicus in an ecological-economic model

W. Jager a,*, M.A. Janssen b, H.J.M. De Vries c, J. De Greef d, C.A.J. Vlek a

a Centre for En6ironmental and Traffic Psychology (COV), Uni6ersity of Groningen, Grote Kruisstraat 2/I,9712 TS Groningen, The Netherlands

b Department of Spatial Economics, Free Uni6ersity, De Boelelaan 1105, 1081 HV, Amsterdam, The Netherlandsc National Institute of Public Health and the En6ironment, Department for En6ironmental Assessment (RIVM-MNV). PO Box 1,

3720 BA, Biltho6en, The Netherlandsd Faculty of Law, Uni6ersity of Groningen, Oude Kijk in’t Jatstraat 26, 9712 EK, Groningen, The Netherlands

Abstract

In mainstream economy, behaviour is often formalised following the rational actor-approach. However, in real lifethe behaviour of people is typified by multidimensional optimisation. To realise this, people engage in cognitiveprocesses such as social comparison, imitation and repetitive behaviour (habits) so as to efficiently use their limitedcognitive resources. A multi-agent simulation program is being developed to study how such micro-level processesaffect macro-level outcomes. Sixteen agents are placed in a micro-world, consisting of a lake and a gold mine. Eachagent’s task is to satisfy its personal needs by fishing and/or mining, whereby they find themselves in a commonsdilemma facing the risk of resource depletion. Homo economicus and Homo psychologicus are formalised to study theeffects of different cognitive processes on the agents’ behaviour. Results show that for the H. psychologicus thetransition from a fishing to a mining society is more complete than for the H. economicus. Moreover, introducingdiversity in agents’ abilities causes the H. economicus on the average to decrease its time spent working, whereas forthe H. psychologicus we observe an increase in the time spent working. These results confirm that macro-levelindicators of sustainability, such as pollution and fish-harvest, are strongly and predictably affected by behaviouralprocesses at the micro-level. It is concluded that the incorporation of a micro-level perspective on human behaviourwithin integrated models of the environment yields a better understanding and eventual management of the processesinvolved in environmental degradation. © 2000 Elsevier Science B.V. All rights reserved.

Keywords: Commons dilemma; Resource; Consumat; Simulation; Multi-agent; Dynamics; Psychology

www.elsevier.com/locate/ecolecon

* Corresponding author: Tel.: +31-50-3636482; fax: +31-50-3636308.E-mail address: [email protected] (W. Jager).

0921-8009/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved.

PII: S 0921 -8009 (00 )00220 -2

W. Jager et al. / Ecological Economics 35 (2000) 357–379358

1. Introduction

To improve our understanding of the world,scientists usually try to capture important as-pects of a real-world system in a model. Scien-tific models allow for experimentation withdifferent assumptions regarding how the real-world system works, thereby improving our un-derstanding of the processes that govern thebehaviour of a real-world system. Scientific mod-els can, in a broad sense, be considered as moreor less formalised mental maps (Rosen 1987,1991). They play a crucial role in research andtheir development often proceeds through iso-morphisms, that is, by hypothesising that twodifferent real-world systems share the same for-mal description. This modelling-by-analogy hasplayed an important role in economic science inan attempt to describe complex economic(sub)system(s). For instance, there are clear iso-morphisms between neo-classical equilibrium the-ory and the laws of chemical equilibrium asformulated by, e.g., Guldberg and Waage (1867),and between theory on the production of goodsand classical thermodynamics. In the physicalsciences, the rather black-box macro-level de-scription of physical systems has been comple-mented with micro-level descriptions in whichsystem elements are identified as separate entitieshaving certain characteristics such as mass andvelocity. In ecological and economic systems theelements have many more degrees of freedomand a much larger variety in their interactions— which makes them complex systems — andhence a modelling tool must be applied that iscapable of dealing with this complexity whilstyielding comprehensible outcomes. Multi-agentsimulation provides a tool that is suitable forstudying how micro-level characteristics affectmacro-level system behaviour.

The growth in computer processing speed wasone of the important conditions for the rapidgrowth of multi-agent simulations to study be-havioural dynamics in social settings. Overviewsof this developing area can be found in Val-lacher and Nowak (1994), Doran and Gilbert(1994), Gilbert and Conte (1995), Conte et al.(1997), Liebrand et al. (1998) and Gilbert and

Troitzsch (1999). Especially within the paradigmof game theory many simulations have been de-veloped that involved the management of a com-mon resource. However, in these simulations theagent rules are usually aimed at maximising out-comes instead of representing real human be-haviour.

In the multi-agent approach applied in thework presented here, agents are equipped withexplicit behavioural strategies that depend on in-formation and expectations about the environ-ment — including other agents. In this waymore behavioural richness can be incorporatedin the simulations than has been done so far.This allows including a larger part of valid in-sights from social science research about humanbehaviour. Thus the multi-agent model gives amore intricate understanding of the system andof the ways in which it could be influenced. Thedisadvantages mirror these advantages: multi-agent models are hard to validate empirically,and they often have to rely on anecdotal evi-dence for their credibility. The reason is that it isdifficult to find unequivocal empirical evidencefor the very micro-level laws that give the mod-els their richness. It may well be that this ’weak-ness’ of our empirical knowledge is an inherentfeature of complex systems (Janssen and DeVries, 1999). An interesting point is to seewhether these simulations give rise to certainmacro-properties (’emergent properties’) whichlink them to the classical macro-approaches insocial and economic science.

In this paper we demonstrate a multi-agentmodelling approach aimed at introducing psy-chological rules that guide the behaviour of ar-tificial agents. These artificial agents, which wehave named ‘consumats’, are being tested in theecological-economic model called ‘Lakeland’.Lakeland includes a renewable fish stock and agold mine as need-satisfying behavioural oppor-tunities. Before we demonstrate the consumats’behaviour in Lakeland, we will first discuss asocial-scientific perspective on resource manage-ment and the common dilemma paradigm. Fol-lowing that, we will dedicate a separate sectionto the consumat model.

W. Jager et al. / Ecological Economics 35 (2000) 357–379 359

2. Human behaviour and renewable resources

Man’s relation with ecosystems is a double-faced one. On the one hand, we depend on ecosys-tems as resources for food, building materials anda healthy environment to live in. On the otherhand, we often plunder and pollute ecosystems asif we were independent from them. This oftenresults in the depletion of natural resources. Thecentral question here is why people bite the handthat feeds them.

Economists were one of the firsts to put re-source depletion on the research agenda. In thelate 18th and early 19th century, Malthus, Ri-cardo and Mill all concluded that scarcity ofnatural resources could lead to diminishing re-turns, and thus to a reduction in economic output(Norton, 1984). According to Malthus (1789),land resources were a limiting factor in feeding anincreasing population. Ricardo (1817) distin-guished different types of land, and argued thatfirst, land of the best quality is used before land ofa lower quality is used, a process leading toincreasing marginal costs of additional land use.According to Mill (1848), resource productivitycould be maintained or even improved if onepromotes technological development. The argu-ments from the early days of economics still holdin the current discussions on the limits to growth.Since our focus is on renewable resources, we willdiscuss this type of resource economics in moredetail.

A renewable resource, such as a fish stock,timber and clean air, has a maximum carryingcapacity depending on the environmental condi-tions (Clark, 1976). Changes in these environmen-tal conditions can affect the carrying capacity ofthe resource. The maximum sustainable yieldfrom a renewable resource is the maximum quan-tity that can be withdrawn from the resourcewithout reducing its size after renewal of theresource (e.g., restoration of the fish stock). Froma classical economic perspective, however, one isnot interested in this sustainable yield, but in thelevel of extraction (e.g., harvest, catch) that max-imises the returns on investments. The optimallevel of extraction is the level where the marginalcosts equal the marginal returns from extraction.

If one considers extraction during a number ofperiods, economists are interested in maximisingthe present value of extraction, that is, the dis-counted stream of costs and benefits from theefforts to extract. Since discounting is dependenton returns from other investments, depletion ofthe renewable resource is likely when reinvestingthe revenue from resource extraction is moreprofitable, e.g., fishermen investing their revenuesin better ships.

When more agents have free access to the re-source, the situation becomes that of a commonproperty. Hardin’s frequently cited article on the‘tragedy of the commons’ explains why peopleoften tend to overexploit collective resources(Hardin, 1968). Here the individual agents are notprimarily concerned with their marginal effect onthe resource, but with the actual returns fromextraction they derive for themselves. This leadsto a higher extraction rate, which increases thechance of resource depletion, but certainly resultsin economic overexploitation (Norton, 1984, p.119). However, collective resources are not alwaysplundered, but are often exploited in a sustainableway, frequently guided by group norms regardingthe harvesting behaviour (e.g., Ostrom et al.,1999). Such situations, in which the behaviourthat is in the individual’s interest is not optimalfrom the group’s aggregate perspective, and viceversa, are called commons dilemmas (Hardin,1968).

The commons dilemma, involving a conflictbetween individual and collective interests, hasintrigued scientists since Machiavelli (1525), whoaddressed this problem in the context of the polit-ical consequences of social (in)equality. Today,the commons dilemma is being used to under-stand the (over)use of exhaustible natural re-sources without property rights. Exhaustibleresources may be renewable, such as fish stocksand timber, and non-renewable, such as oil wellsand ore. Mainstream economists argue that thereexist several rationalities that may guide man’suse of renewable resources. For example, types ofrationality lead to individual rationality, the Nashequilibrium, Pareto efficiency and strong equi-librium (Dasgupta and Heal, 1979). Economicresearch clarifies that certain rationalities, when


applied to a renewable resource, may yield out-comes that are less optimal and sometimes evendisastrous. For example, individual rational be-haviour may result in uneconomic harvesting be-haviour, even if such behaviour does notjeopardise the resource itself (Norton, 1984).

Much economic research is aimed at the condi-tions under which certain types of rationality aredominating, and what types of measures are feasi-ble to affect this domination. In this research,economists employ the ‘rational actor’ approach.This implies that individual choice behaviour isbeing modelled following the assumption thatpeople have perfect knowledge and try to optimisetheir outcomes. Research using this rational-actorapproach is usually focused on finding the condi-tions for an optimal resource extraction rate, pro-vided that the actors employ an individualoptimising strategy. Moreover, it can be calcu-lated what governmental measures are necessaryto attain an optimal extraction rate given that allagents are following such an optimising strategy.This approach provides an essential back-curtainagainst which the fallacies of actual economicbehaviour may be perceived.

Where economics is generally a normative sci-ence studying the optimal way of allocating scarceresources, (social) psychology is a more descrip-tive science studying the actual decision making ofhuman actors. Luce and Raiffa’s Games and Deci-sions (Luce and Raiffa, 1957) awakened a fascina-tion for the conflict between individual andcollective interests in many social scientists alsooutside the economic tradition. Especially socialpsychology questions what factors affect the ac-tual harvesting behaviour of people confrontedwith renewable and non-renewable common re-sources. Combining a social-psychological viewon behaviour with the economical principles thatapply to renewable resource management mayadd to the understanding of the discrepanciesbetween optimal and actual resource-usebehaviour.

A major assumption in social psychology is thatpeople do not always optimise their outcomes, butoften engage in satisficing behaviour (Simon,1976). Because people have limited cognitive re-sources, much of their daily routines are auto-

mated. Here, optimising is not focused exclusivelyon behavioural outcomes, but also on managingcognitive resources. Automating behaviour, forexample, prevents people from spending hours insupermarkets deliberating over the various brandsof products they can buy. Instead, people oftenhabitually repeat their originally deliberatechoices for as long as the outcomes are satisfying.However, a current habit may yield far fromoptimal outcomes because new, better be-havioural opportunities may have been intro-duced. Especially in the context of depletingrenewable resources, it may be so that people whoconsume in a habitual manner may be unaware ofthe developments in the resource size, which maycause under- or over-harvesting.

A second major assumption of social psychol-ogy is that people often learn about new attractivebehaviours by using information about other peo-ple’s behaviour. This so-called social processingcan also be understood as a strategy to economisetheir cognitive efforts, and thus it fits in a broaderperspective on optimising. A major finding is thatpeople tend to engage in social processing whenthey are uncertain (Festinger, 1954). This is rele-vant for understanding how people use renewableresources such as ecosystems, because these areoften quite complex systems, and their dynamicsmay yield more or less surprising outcomes. Forexample, a fish stock may manifest large oscilla-tions related to variations in factors such as watertemperature, fish-harvest, the availability of food(e.g., algae) and pollution (e.g., Walker and Liv-ingstone, 1992). As a consequence, people mayexperience different levels of uncertainty regard-ing the state of the resource, which may affecttheir harvesting behaviour.

Many social–psychological laboratory experi-ments have been aimed at unravelling the factorsthat affect people’s harvesting behaviour. Here,subjects are confronted with a common resourceand they have to decide how much to take fromthat resource. Because this experimental gamemay last for an extended number of rounds in aprolonged period of time, people have to take thelong(er)-term effects of their behaviour on theresource into account. This can be easily relatedto the concept of sustainability that is discussed a


lot within the environmental sciences. Typicalfounders in this experimental tradition wereJerdee and Rosen (1974), Rubenstein et al. (1975)and Brechner (1977). Group factors, personal fac-tors and resource characteristics have been experi-mentally found to affect subjects’ harvestingbehaviour in a resource dilemma.

Group factors that have been found to increaseharvesting are, e.g., a larger group size (e.g. Foxand Guyer, 1977), a low identifiability of theindividual behaviour (e.g., Jorgenson and Pap-ciak, 1981) and a weak group identity (e.g.,Brewer, 1979; Edney, 1980). Personal factors thathave been found to increase harvesting are, e.g., aweak belief that personal restraint is essential tomaintain the resource (e.g., Jorgenson and Pap-ciak, 1981; Samuelson et al., 1984), high uncer-tainty regarding the behaviour of others, (Messicket al., 1988), and an individual or competitivesocial value-orientation (Messick and McClin-tock, 1968; McClintock, 1978). Several resourcecharacteristics also affect the harvesting behaviourof people. Obviously, the relevant resource dy-namics are important. People will harvest morewhen the resource size is larger and the larger isits regenerative capacity. Several researchersfound that increased environmental uncertainty(see above) also caused people to harvest more(Messick et al., 1988; Suleiman and Rapoport,1989; Budescu et al., 1990; Rapoport et al., 1992;Hine and Gifford, 1996; Wit and Wilke, 1998).Effects of uncertainty have even been demon-strated to affect the harvesting behaviour of expe-rienced fishermen managing a simulated fish stock(Moxnes, 1998).

Whereas experimental laboratory research hastaught us a lot on the factors that affect be-haviour in commons dilemmas, these results can-not be translated directly towards real-lifesituations. Most psychological experiments areconducted within the hour, whereas real-lifedilemmas may exist for decades or even longer.Moreover, the choices that people make whileplaying an experimental game usually have nofar-reaching consequences for their lives. In real-life dilemmas, however, the choices one makesmay determine one’s quality of life to a far extent.

Field experiments and observations mayprovide the data that allow for inferring be-haviour-determining factors in real-life situations.Field experiments, however, provide data of lim-ited validity because they are based on limited(quasi-)experimental variations applicable duringrelatively short time periods. Observational datausually do not allow for drawing causal inferencesbecause the complex relations between the vari-ables and environmental conditions are not wellknown.

Computer simulation of artificial agents in acommons-dilemma situation offers a tool to ex-periment with more real-world conditions andwith long time-series. Despite the fact that com-puter simulations are usually quite simple in com-parison to the real world, the dynamics beingexplored may help interpreting real-life dilemmas.As such, simulation is an approach which, incombination with other methodologies, con-tributes to the understanding of why people incommons dilemmas behave like they do and whatstrategies may be viable in altering less sustainablebehaviours.

Within game theory, many simulations of com-mons dilemmas have been developed. The com-puter tournament for the most viable strategy,being organised by Axelrod (1980a,b, 1984), wasan important stimulus for this kind of research.Many researchers developed strategies for au-tomata, such as win-stay, lose change, or win-stay, lose-defect. These strategies were beingapplied in automata playing iterative games so asto explore the dynamics of resource use (e.g.,Liebrand and Messick, 1996). A very good strat-egy appeared to be cooperating unless the otherautomata defected in the previous round. Thisstrategy is known as Tit-for-Tat. Later, moresophisticated rules were developed, e.g. by equip-ping the automata with a memory for the be-haviour of other automata (Lindgren andNordahl, 1995; Molander, 1985; Nowak and Sig-mund, 1992), or by equipping the automata withevolutionary strategies (genetic algorithms), whichallowed them to adapt their choice behaviour tothe behaviour of the other agents (e.g., Axelrod,1987; Macy, 1996). However, most game theoreti-cal simulations were primarily focussed on devel-


oping optimal strategies instead of trying to de-velop strategies that reflect human decision-mak-ing. To increase the behavioural richness in thissimulation paradigm it is necessary to representthe cognitive processes that guide the harvestingbehaviour of real people in agent rules.

In real-life dilemmas, cognitive processes thatappear to be relevant for understanding how re-newable resources are being managed, refer todeliberate choice behaviour, social comparison,imitation and habit formation, respectively. Manybehavioural theories are available to guide thedevelopment of agent rules involving such cogni-tive processes, e.g., theories on deliberate be-haviour, normative conduct, social comparisonsand learning processes. Because the type of cogni-tive processing a person engages in depends onthe situation it encounters, it is necessary tospecify under what conditions, which theory-based rule will guide an agent’s behaviour. Totake account of a variety of essential behaviour

determinants and mechanisms, a conceptualmodel for consumer behaviour has been devel-oped that integrates relevant behaviour theories(Jager et al., 1997; Jager, 2000). This conceptualmodel has been used to develop a comprehensiveset of theory-based agent rules (Jager et al., 1999;Jager 2000), which we have named the consumatapproach.

3. The Consumat approach

Many behaviour theories each explain parts ofthe various processes that determine consumerbehaviour. Due to this, social psychology is some-times said to be in a pre-paradigm state; accord-ing to some scholars there is a need for ameta-theory of human behaviour (Vallacher andNowak, 1994). Such a meta-theory shouldprovide a framework to integrate simplified ver-sions of relevant expert-theories of behaviour. The

Fig. 1. The conceptual model of consumer behaviour (source: Jager, 2000).


consumat approach as introduced by Jager et al.(1999) is based on a comprehensive conceptualmodel of consumer behaviour, which could be thebasis for such a meta-theory. This conceptualmodel is schematised in Fig. 1.

The driving forces at the collective (macro-) andthe individual (micro-)level make up the environ-mental setting for consumer behaviour. The collec-tive level refers to technical, economical,demographic, institutional and cultural develop-ments. At the individual level there are the needsof consumers, the opportunities to be consumed,and individual abilities for consuming the oppor-tunities. Furthermore, consumers have a certaindegree of uncertainty regarding the outcomes ofbehaviour.

Dependent on the level of need satisfaction,which is based on the satisfaction of differentneeds, and the degree of uncertainty, certain cog-nitive processes determine the choices for con-sumption along the dimensions social processingand reasoned beha6iour. Consumats having a lowlevel of need satisfaction and a low degree ofuncertainty are assumed to deliberate, that is,assessing the consequences of all possible decisionsgiven a fixed time-horizon in order to maximiselevel of need satisfaction. Consumats having a lowlevel of need satisfaction and a high degree ofuncertainty will engage in social comparison. Thisimplies the comparison of its own previous be-haviour with the previous behaviour of consumatshaving somewhat similar abilities, and selectingthat behaviour which yields a maximal level ofneed satisfaction. When consumats have a highlevel of need satisfaction, but also a high level ofuncertainty, they will engage in the imitation of theconsumptive behaviour of other, somewhat similarconsumats. Finally, consumats having a high levelof need satisfaction and a low level of uncertaintyhabitually repeat their previous behaviour. Whenconsumats engage in reasoned behaviour (deliber-ation and social comparison), they will update theinformation in their mental map, which serves asa memory to store information on abilities, oppor-tunities, and characteristics of other agents. Fromthe consumption of opportunities, a new level ofneed satisfaction will follow. Uncertainty in theconsumats is defined as the difference between the

actual and the expected level of need satisfaction.When this difference exceeds the uncertainty toler-ance level, the consumat will engage in socialprocessing; below this uncertainty tolerance levelthe consumat will engage in individual processing.Consumption leads to changes in abilities, oppor-tunities and the social and physical environment,which will affect consumption in subsequent timesteps.

Based on the conceptual model, a multi-agentcomputer simulation model has been developed.The consumats are equipped with various needs,and they may consume opportunities to satisfythese needs. The specific level of consumptionfrom a specific set of opportunities j=1..J is beingnotified as xj, which stands for, e.g. a certainnumber of hours spent fishing. For opportunity jand a need i from a specific set of needs i=1..I,the level of needs satisfaction N at time t isformulated as a diminishing marginal utility func-tion, which implies that the more of an opportu-nity is being consumed, the less utility one getsfrom consuming yet another unit of that opportu-nity:

Ni,t=1−exp(−a�xj) (1)

with parameter a indicating the sensitivity of Ni

for the consumption of opportunity j at time t.The total level of need satisfaction of a consumatis a weighted multiplication of the satisfaction ofthe set of needs i=1..I.

Nt=N1,tg2�N2,t

g2�…�Nn,tgn (2)

It is postulated that there is a critical value forN, NMIN, below which a consumat is dissatisfiedand may change its behavioural mode. Because ofthe multiplicative need-satisfaction function, avery low level of need satisfaction for any singleneed may result in a low overall N and hence causedissatisfaction.

The consumat is equipped with abilities thatrefer to, e.g., financial means, time perspective(time-horizon used in calculations) and harvestingcapacity. The consumat is also equipped with amental map to memorise previous behaviours ofitself and others, and to memorise the need satisfy-ing potential of various opportunities. The con-


sumat k experiences uncertainty U at time t pro-portional to the absolute difference between theexpected consumption of opportunity j, E(xj), att−1, e.g. during harvest, and its actual consump-tion at t−1:

Uk,t=ABS(E [xj,t−1]−xj,t−1) (3)

A maximum level of uncertainty tolerance, UT,is postulated: if the uncertainty exceeds this value,the consumat will become uncertain. Dependingon the consumat’s level of need satisfaction anduncertainty level, it may engage in four differentcognitive processes: deliberation, social compari-son, repetition and imitation; see below. It isimportant to note that a distinction is made be-tween social processing (social comparison andimitation) and social information (e.g., relativefinancial budget, the behaviour of other con-sumats). For example, consumats may individu-ally deliberate on how to maximise theiroutcomes, thereby making assumptions on thebehaviour of the other consumats.

3.1. Repetition (consumat is satisfied and certain)

Repetition stands for automatic individual pro-cessing, and relates to classical and operant condi-tioning theory (Pavlov, 1927; Skinner, 1953). Amajor assumption in these theories is that be-haviour is reinforced when the outcomes are posi-tive. A consumat engages in repetition when it issatisfied (Ni,t]Nmin) and certain (Uk,t5UT). Re-peating implies that the consumat does not updateits mental map. It will just repeat the previousbehaviour (opportunity consumption) in order toremain satisfied.

xj,t=xj,t−1 (4)

Thus, the consumat is motivated to consumethe previously consumed opportunity. Only whenit appears that the behavioural control over thisopportunity has dropped to zero, indicating thatthe consumat lacks sufficient resources (e.g.,money) to consume the opportunity in question,the consumat will switch towards deliberating tofind an opportunity that is both satisfying andfeasible.

3.2. Deliberation (consumat is dissatisfied andcertain)

Deliberation stands for reasoned individualprocessing, as conceived by decision and choicetheory (Janis and Mann, 1977; Hogarth, 1987)and theory of reasoned action (Fishbein andAjzen, 1975; Ajzen, 1991; Ajzen and Madden,1985). A central assumption of these theories isthat people assess the possible outcomes of theirspecific behavioural opportunities o, and choosethe behaviour that maximises these expected out-comes. Consumats engage in deliberation if thelevel of need satisfaction is below the minimumthreshold level (Ni,tBNmin), and when they arecertain (Uk,t5UT) about behavioural outcomes.Deliberating starts with updating the mental map.This updating implies that information is gatheredregarding the need-satisfying capacities of oppor-tunity consumption, the resource demands of op-portunities (e.g., financial costs), the consumat’sown abilities and the previous behaviour of otherconsumats. This information is being used tocalculate the expected outcomes in terms of levelof need satisfaction and the behavioural controlover all these opportunities. This calculationyields a maximum expected level of N for aparticular set of feasible opportunities.

xj,t=maxNi(x),x�X (5)

In the calculations the consumat uses a fixedtime horizon. Outcomes that fall beyond the timehorizon are not taken into consideration, and thusdo not affect current behaviour. The consumatwill be motivated to consume the specific oppor-tunities that jointly yield the highest perceivedneed-satisfaction that is feasible in terms of be-havioural control.

3.3. Imitation (consumat is satisfied anduncertain)

Imitation stands for automatic social process-ing, and relates to social learning theory (Ban-dura, 1977, 1986) and the theory of normativeconduct (Cialdini et al., 1991). Both theories sug-gest that imitating others’ behaviour may bebeneficial. Social learning theory states that


watching another person being rewarded for un-derstandable and reproducible behaviour may re-sult in the imitation of that behaviour. The theoryof normative conduct states that people often dowhat others would approve of them to do. Aconsumat engages in imitation if it is satisfied(Nj,t]Nmin) and uncertain (Uk,t\UT). When theconsumat engages in imitation, it will read themental map and recall the consumat that func-tioned most recently as a comparison-consumat.It will do what this consumat s did in the previoustime-step. The consumat is thus motivated toconsume whatever the other similar consumat hasbeen consuming in order to keep satisfied.

xk,i,t=xs,i,t−1 (6)

in which xs,i,t−1, is the previous behaviour of asimilar consumat and Xk,i,t−1 is the previous ownbehaviour. Also the behavioural control over thisbehaviour is being calculated. Only when it ap-pears that the behavioural control over this op-portunity is below zero, the consumat is notcapable of consuming this opportunity, and con-sequently will switch towards social comparison(see below).

3.4. Social comparison (consumat is dissatisfiedand uncertain)

Social comparison stands for reasoned socialprocessing, and relates to social comparison the-ory (Festinger, 1954; Masters and Smith, 1987)and theory of reasoned action including socialnorms (Fishbein and Ajzen, 1975). Social com-parison theory states that people have a drive toevaluate the behaviour of similar others whenthey feel uncertain about which behaviour to per-form. Moreover, the theory of reasoned actionstates that the behaviour of other people is oftenconsidered as a norm. A consumat engages insocial comparison if it is dissatisfied (Ni,tBNmin)and uncertain (Uk,t\UT). While engaging in so-cial comparison, the consumat first will update itsmental map. Then it will observe the previousconsumptive behaviour of the other consumatshaving somewhat similar abilities. When otherconsumats’ abilities differ no more than a certainpercentage from its own abilities, those consumats

are assumed to be similar, and thus comparable.The consumat will calculate the expected out-comes for reproducing the opportunity consump-tion of the other consumat. The consumat will bemotivated to copy either the other consumat’sbehaviour, or its own previous behaviour, de-pending on which of the two yields the highestneed-satisfaction.

xk,i,t=maxNi(xk,i,t−1,xs,i,t−1),xs,i,t−1�Xs,i,t−1 (7)

in xs,i,t− l is the previous behaviour of a similarconsumat, Xk,i,t−1 is the previous own behaviourand Xs the set of behaviours of similar consumatsat t−1. Also the behavioural control over thisbehaviour is being calculated.

3.5. Validation of the consumat approach

As was noted in the introduction, it is difficultto validate a multi-agent simulation because ofthe complexity of the systems that are being simu-lated. We tried to validate the consumat approachon two levels: the micro-level rules and the result-ing behavioural dynamics.

Regarding the micro-level rules, the validationrests in the fact that the consumat rules are beingdeveloped on the basis of validated psychologicaltheories. Of course, a severe reduction took placein order to develop the cognitive processing rules.Moreover, the rules that determine when whichcognitive processing rule is being followed arequite rudimentary because no validated integratedbehavioural model exists to validate these rules.However, the formalisation of relevant be-havioural processes in the consumat is expected toincrease the behavioural richness in simulationmodels in comparison to formalising agents thatexclusively engage in outcome-maximisingprocesses.

The validation of the resulting behavioural dy-namics resides in the capacity of the consumatapproach to replicate empirical findings. Here, weare confronted with a validation problem becauseresource management experiments using real peo-ple have not collected data on the cognitive pro-cesses people engaged in. The validation optionthat remains open is to check if the simulationresults are at least in accordance with the effects


that have been found in empirical studies. There-fore, we decided to confront our consumats with asimple renewable resource (Jager et al., 1999). Weconcentrated on the effect of uncertainty becausethis had been investigated by several researchersusing human subjects, yielded clear results, andappeared to be relatively simple to formalise inthe consumat approach. Just as in empirical stud-ies, we found that an increase in uncertaintycaused consumats to harvest more from a com-mon resource. These results at least increase theface-validity of the consumat approach.

The consumat simulation of a common re-source revealed three dynamical processes thatexplained why increased environmental uncer-tainty provokes over-harvesting. These are, re-spectively, the optimism effect, the imitation effectand the adaptation effect. The optimism effectholds that deliberating consumats, when con-fronted with a positive fluctuation in resourcegrowth, are more likely to develop an over-har-vesting habit because over-harvesting increasesthe chances of the consumat being satisfied in theshort run. The imitation effect implies that con-sumats, when uncertain and satisfied, are likely toimitate the behaviour of the other consumat, evenwhen this behaviour is less optimal than one’sown previous behaviour. The adaptation effectholds that no new behavioural opportunities arebeing adopted during social processing, and as aconsequence the consumats are not capable ofadapting their behaviour to changing circum-stances, such as a serious depletion of the re-source. An essential aspect of these process effectsis that they are originating from the differentbehavioural processes that the consumats engagein. Consequently, the consumat-approach, intro-ducing a broader perspective on rational be-haviour, allows for the expression of relevantconsumer dynamics, such as habit formation andsocial learning, in ecological economic models.Empirical study of these process effects is impor-tant both for a better understanding of the be-havioural dynamics of resource management aswell as for validation of the consumat approach.In the following section we will present varioussimulation experiments using consumats in anecological economic model.

4. Fish, gold, and the happiness of consumats

To experiment with simulated behaviour in anecological–economic model, we place the con-sumats in a ‘micro-world’ called ‘Lakeland’ (DeGreef and De Vries, 1991). Lakeland consists oftwo natural resources: a fish stock in a lake and anearby gold mine. The lake is being modelled as asimple ecological system of fish and shrimps. Theintroduction of consumats in an ecological–eco-nomic model implies that different behaviouralprocesses, such as social comparison and imitativebehaviour, underlie consumats’ harvesting be-haviour. We place sixteen consumats in Lakeland,and these consumats catch fish from the lake tosatisfy their need for food. Moreover, they mayalso export fish, and the returns can be spent onassets such as luxury goods. Import of fish is alsoallowed. If the gold mine is opened, consumatsmay also dig for gold. The money that is earnedby mining can be spent on fish imports and/or onluxurious assets. The pollution that is caused bymining is reducing the carrying capacity of thelake for the fish and shrimp populations. Theconsumats determine how they allocate their timeon leisure, fishing and mining. They are equippedwith certain abilities for fishing and mining, andwant to satisfy their four needs: leisure, identity,subsistence and freedom. These needs, which aresupposed to play an important role in this case,are selected out of a larger set of needs as de-scribed by Max-Neef (1992). We assume that thesatisfaction of the leisure need relates to the shareof the time spent on leisure. The identity need issatisfied by the relative amount of money theconsumat owns in comparison to consumats withsimilar abilities. The subsistence-need relates tothe consumption of food. The freedom need isassumed to be related to the absolute amount ofmoney the consumat owns, which can be spent onwhatever assets the consumat prefers.

In Fig. 2 we depict the variables used in thesimulation model in regard to the conceptual be-havioural model as depicted in Fig. 1.

To study how the different possible behaviouralprocesses may affect the interactions with thesimulated environment, we decided to focus onexperimental manipulations of consumats’ sup-


Fig. 2. The variables used in the simulation model.

posed tendency to engage in particular be-havioural processes. These experiments were per-formed in two set-ups of Lakeland, one in whichonly fishing is allowed, and a second in whichmining is allowed besides fishing. First we willdescribe the results we obtained with Lakelandwhere only fishing is allowed, and in a next sec-tion we will present results of introducing miningin Lakeland.

4.1. Fishing in Lakeland

In the first simulation experiments the set-up ofLakeland does not allow consumats to dig gold,and thus they fully depend on their fishing capac-ities to satisfy their needs. Varying the minimallevel of need satisfaction and the level of uncer-tainty tolerance allows for the specification ofconsumats that engage in one or more of the fourbehavioural processes distinguished in Fig. 2. Two

types of consumats have been formalised to showhow different types of behaviour affect the man-agement of the fish stock. The first type is rela-tively quickly satisfied, but also quickly uncertain(Nmin=0.05, UT=0.05). This leads this type ofconsumat to engage in all four behavioural strate-gies. This consumat, having a large degree offreedom regarding the behavioural strategies, willcorrespondingly be denoted as the Homopsychologicus.

The second consumat type is virtually neversatisfied or uncertain (Nmin=0.95, UT=0.95).This causes this type of consumat to engage exclu-sively in outcome-optimising deliberation. Thisconsumat can be considered as a simple version ofthe optimising agent in economic models, and itwill consequently be denoted as the Homo eco-nomicus. The H. economicus is self-supportive, i.e.not learning from the behaviour of other con-sumats nor making explicit assumptions about


the decisions of others. The consumats employ atime horizon (cognitive ability) of 10 time-stepswhen engaging in reasoned processing (delibera-tion and social comparison). Moreover, the con-sumats have a fishing ability of 6600, indicatingthat they are capable of catching 6600 units of fishper time-step when the fish stock is on its carryingcapacity level (the maximum number of fish-unitsis about 23 000). Factors that affect the actualcatch are the fish density, the fish growth function(in our model about 2.3, indicating that at eachtime-step the number of fish is multiplied by 2.3for as long as the maximum is not reached) andthe length of the fishing season (in our model 80%of the time). Consequently, we have a very viablefish stock, which renews itself very quickly, andhence is quite difficult to deplete. To introduceenvironmental uncertainty in the model, astochastic function in the fish catch has beenformalised.

For both experimental conditions (H. economi-cus and H. psychologicus) 100 model runs havebeen performed, for which the average outcomeswill be presented. Each experiment involved six-teen consumats with identical fishing abilities in-habiting Lakeland. Thus, all consumats wereeither H. economicus or H. psychologicus, depend-ing on the experimental condition. The consumats

start with an equal financial budget. Because theyare depending on each other regarding the man-agement of the resource, the consumats are actu-ally caught in a commons dilemma.

In reporting the results we will focus on a fewessential variables. First, the behaviour and be-havioural processing of the consumats will beobserved. Second, the developments in the fishstock will be considered. Third, the developmentsin the need satisfaction of consumats will beobserved. Finally, attention will be given to thedevelopments in the financial budget of the con-sumats, which provides a material index of wel-fare in the model.

Each simulation experiment involves 50 time-steps, each time-step representing a year in whichconsumats must decide how much time to spendon fishing and leisure in order to satisfy theirneeds.

4.1.1. ResultsLooking at the behavioural processes the

consumats engage in, it appears that the H. psy-chologicus engages in all four behavioural strate-gies, as was intended. Fig. 3 shows for time-step 1 to 50 the proportion of time each be-havioural process is engaged. These results are an

Fig. 3. Proportional distribution of the four behavioural processes for the homo psychologicus.


Fig. 4. The development of the fish stock in time for the twoconsumat-type conditions.

Fig. 5. The proportion of time spent fishing for the twoconsumat-type conditions.

financial budget during the first 35 time-steps.However, starting from the moment that the fishstock becomes depleted, the financial budget of H.economicus quickly drops to zero. For the H.psychologicus a decline of the initial financial bud-get can be observed until a relatively stable budgetis reached that is purely earned by fishing.

4.1.2. Homo psychologicusAs was shown in Fig. 3, during the 100 experi-

mental runs the H. psychologicus engages in allfour behavioural strategies. Consumats startfishing for 60% of the time (initial setting). Anoccasional bad catch may yield a lower financialbudget in comparison to other consumats. As aconsequence, this consumat’s level of need satis-faction for ’identity’ drops below the critical level,causing it to engage in reasoned processing (delib-eration or social comparison). When deliberating,this consumat will increase its time spent fishing

average outcome of 100 simulation runs under thecondition of all consumats being H. psychologicus.

The sudden shift from much repetition to muchdeliberation at t:18 is caused by the fact that theinitial financial budget that the consumats use tobuy luxury goods or to import/buy fish has de-pleted around that time. This worse financial situ-ation causes the consumats to be dissatisfied moreoften, causing them to engage in deliberationmore often. Because the consumats are occasion-ally uncertain and hence engage in social process-ing (imitation and social comparison), thisdecrease in need-satisfaction also causes the rela-tive high proportion of imitation before t:18 tochange towards social comparison (see Fig. 3).The H. economicus engaged purely in deliberation,as it was defined to do.

Fig. 4 shows the development in time of the fishstock for these two conditions (100 simulationruns per consumat-type condition). It can be ob-served that the fish stock depletes for the H.economicus, and remains at a high level for the H.psychologicus.

These effects can be attributed to the propor-tion of time spent fishing, which is graphicallydepicted in Fig. 5. H. psychologicus slowly in-creases its fishing to 0.7 of the time, whereas H.economicus more rapidly increases its fishing to0.8 of the time. The explanation of these effectswill be done below in two separate sections for theH. psychologicus and H. economicus.

Differences in time spent fishing have conse-quences for the financial budget of the consumats.Fig. 6 shows that the higher proportion of timespent fishing by H. economicus yields a higher

Fig. 6. The financial budget for the two consumat-type condi-tions.


to about 80% (while the satisfied consumats re-main at 60%) to increase its financial budget. Thisis mostly a satisfactory action, and as a conse-quence this consumat will continue to processautomatically. This implies repetition of its ownprevious behaviour, or imitation of the behaviourof similar consumats. The latter may cause theconsumat to decrease its time spent workingagain. Other satisfied consumats may simply imi-tate the increase in time spent working withoutever engaging in reasoned processing. However,most of them remain fishing for 60% of the time.In some runs, however, the number of consumatsthat imitate the increase in time spent fishing is solarge that the resource becomes seriously de-pleted. This causes the slightly downward devel-opment of the fish stock beyond step 36 that canbe seen in Fig. 4. On the average, the proportionof time the consumats spend fishing increases toabout 67%.

The consumats that increase their time spentfishing will be more likely to catch a surplus offish that they can sell on the market. This allowsthem to earn some money for luxury products,which satisfies their need for freedom. Moreover,their level of need-satisfaction for identity is rela-tively high because they have more money thanthe consumats that did not increase their timespent fishing. However, the increase in time spentfishing is not that large that their level of need-satisfaction for leisure drops significantly. Finally,they catch enough fish to satisfy their need forsubsistence. Because the needs of the consumatsthat increased their time spent fishing are stayingat a sufficient level, they experience a high generallevel of need satisfaction, causing them to remainprocessing in an automatic manner. This impliesrepeating their own previous behaviour, or imitat-ing the behaviour of similar consumats.

The consumats that did not increase their timespent fishing in the beginning (say, the first 15time-steps) will catch somewhat less fish, using itexclusively for their own consumption. Not sellingfish on the market implies that their financialbudget declines, which causes the over all declineof financial budget for the H. psychologicus condi-tion (Fig. 6). As a consequence, these consumatscannot afford to buy luxury goods at a given

moment in time. Because of that, their need forfreedom decreases below the critical level aroundt=20, causing them to deliberate and sociallycompare regarding better opportunities. However,often they calculate that increased working mayincrease their level of need satisfaction for free-dom, but at the cost of their level of need satisfac-tion for leisure and identity. Despite the fact thatthey remain dissatisfied as regards freedom, theydo not change their behaviour because there is noother behavioural opportunity yielding a higheroverall level of need satisfaction.

In most cases the consumats have a reasonablelevel of need satisfaction for subsistence, identityand leisure, and a low level of need satisfactionfor freedom. In about half of the simulation runsa subgroup of consumats increases their timespent fishing in an early stage, resulting in ahigher level of need satisfaction for freedom in thelong run.

To obtain an indication of distribution of in-come, Lorenz (1905) and Gini (1912) developedincome-disparity indexes. The Gini index (Gini,1912) is a measure for distributional equality, andprovides a macro-level indicator of economicequality in a society. In the simulation experi-ments, the Gini index is being used to indicatefinancial equality. A Gini index of 1 denotescomplete financial equality, a Gini index of 0complete financial inequality. In the experimentswith Lakeland the consumats all start with anequal financial budget, which implies a Gini indexof 1.0. In the first 20 time-steps a drop to about0.20 was observed. After this sharp decline in theGini index, in about half of the simulation runsan increase towards 1.0 can be seen, whereas theother half shows an irregular pattern, oscillatingbetween 0.20 and 0.30. This oscillating patternoccurs in the simulation runs where a group ofconsumats increased their harvesting behaviour inthe early time-steps because they experienced ‘badluck’ regarding their harvest. Moreover, someconsumats not experiencing this ‘bad luck’ maystill increase their time spent fishing because theyimitate the consumats that had ‘bad luck’. Espe-cially this last group of consumats will earn a lotof money. Because the consumats that increased


their time spent fishing earn more money than theother consumats, the Gini index remains low,indicating distributional inequality.

4.1.3. Homo economicusH. economicus, which is never satisfied and

never uncertain, exclusively engages in delibera-tion. The consumats in this condition immediatelyincrease their time spent fishing to maximise theiroutcomes. Despite the fact that the consumats,having a time horizon of 10, anticipate the col-lapse of the fish resource, they continue fishingintensively. Around time-step 40, the fish stock isdepleted and the level of need-satisfaction of theconsumats drops to 0.0. This situation typicallyresembles the tragedy of the commons, showingthat behaviour following individual rationalitymay lead towards a collective disaster. The con-sumats implicitly ‘expect’ that all other consumatswill continue their over-harvesting. Consequently,each consumat ‘realises’ that their individual be-haviour hardly affects the depletion of the re-source. Following the principles of individualrationality, all consumats consider a high harvest-ing level as optimal, and consequently they do notsucceed in reducing their harvest to a more sus-tainable level. This situation resembles a sort ofself-fulfilling prophecy, as the fish stock depletesbecause the consumats expect it to deplete. Thisoutcome is a typical example of what happens ifno property rights exist on an exhaustible re-source and the fishermen behave in an own-out-come maximising manner.

Using the same decision rules for a single‘meta-actor’ with needs and abilities that are six-teen times as large as those of each of our sixteenconsumats would result in the meta-actor decreas-ing its harvesting now so as to maximise itsoutcomes over the forthcoming ten time-steps.This is due to the fact that it has full control overthe fish stock because it is not being confrontedwith fifteen other consumats.

In the condition of the H. economicus the con-sumats earn an almost equal amount of moneyfor as long as the fish stock is not depleted,resulting in a Gini index of about 0.97. When thefish stock is depleted at around t=40, and con-sumats’ income has dropped to 0, some con-

Fig. 7. Time spent fishing and mining for the two consumat-type conditions for t=1 to t=50.

sumats will run out of money earlier than otherconsumats do, causing the Gini index to drop to alevel of 0.6. Very quickly all consumats havefinished up their finances, and this equal financialsituation for all consumats obviously yields a Giniindex of 1.0.

4.1.4. No 6ariation in fish catchSome variations of the default case have been

performed to assess the sensitivity of several cru-cial assumptions. In the previous experiments, theharvest of a single consumat at a given time-stepwas partly dependent on chance, just like anyreal-world fisherman may have lucky days andbad days. In the present condition, all randomvariation in the fish catch of the consumats hasbeen eliminated. For the rest, the previous experi-mental consumat-type conditions (H. economicusand H. psychologicus) are being replicated. Re-moving random variation now causes the H. psy-chologicus to be certain all of the time, whichwould cause it to engage exclusively in individualprocessing (deliberation and/or repetition).

Because the H. psychologicus does not experi-ence occasional ‘bad luck’ regarding the harvest,its level of need satisfaction remains at a satisfac-tory level. Consequently, the H. psychologicus en-gages only in repetition, fishing for 60% of thetime instead of increasing the time spent fishing toabout 67% as found in the previous experimentthat included a stochastic function in the fishcatch. These results, showing an increase in har-vesting when uncertainty increases, are in accor-dance with the results obtained with a simpleresource (Jager et al., 1999). The H. economi-


Fig. 8. The fish stock for the two consumat-type conditions fort=1 to t=50.

Fig. 10. The financial budget for the two consumat-typeconditions.

cus, however, did not alter its harvesting be-haviour in response to removing all random varia-tion in the fish catch. This is due to the fact thatthe H. economicus has a high tolerance for uncer-tainty (UT=0.95), and thus is less sensitive tovariations in the harvest.

4.2. Introducing mining in Lakeland

In the next series of experiments the set-up ofLakeland allows the consumat to engage in min-ing starting from the first time-step. Consumatsthus can alternately fish and mine to satisfy theirneeds. Besides a fishing ability of 6600 (see previ-ous section), the consumats have a mining abilityof 100, indicating they are capable of harvesting amaximum of 100 units of gold per time-step. Itdepends on the market dynamics (e.g. world fishprice versus local fish price) how much fish theycan buy for a unit of gold. As a starting condi-tion, the consumats are fishing for 60% of the

time, and they do not yet engage in mining. Forthe rest the design of two consumat-type condi-tions as used in the first series of experiments isbeing replicated. Because the mining of gold in-duces pollution of the lake, which in its turnaffects the fish stock, data on the development ofpollution concentration in the lake will also bepresented.

4.2.1. ResultsFig. 7 shows the time spent fishing and mining

for H. economicus and H. psychologicus, respec-tively. Again, the results involve 100 simulationruns per consumat-type condition. It can be seenthat H. economicus very quickly switches to min-ing, but instead of a smooth transition an oscillat-ing pattern between time spent fishing and miningcan be seen. H. psychologicus shows a slow in-crease in mining until at about t=20 the transi-tion is completed.

Looking at the consequences for the fish stock(Fig. 8), first it can be observed that for h. eco-nomicus the fish stock does not fully depletewithin fifty time-steps, as it did in the previouscondition (Fig. 4).

The introduction of mining implies that theconsumats spent less time fishing. Because H.psychologicus spends less time fishing than H.economicus, the latter depletes the fish stock toa larger extent. However, H. psychologicus de-pletes the fish stock at a higher rate than in theprevious condition, despite the fact that it spendsless time fishing. A major reason for this is thatthe slow but persistent move towards mining in-creases the pollution that causes the fish stock to

Fig. 9. The pollution concentration for the two conditions fort=1 to t=50.


decrease (Fig. 9). This further accelerates thetransition to mining, as the fish catches per hourkeep decreasing.

Due to the mining, the consumats are capableof earning more money than in the previouscondition without mining. As Fig. 10 shows, thisyields a higher financial budget for both types ofconsumats than in the previous experiment (Fig.6). Continually deliberating, H. economicussucceeds in attaining a higher financial budgetthan H. psychologicus.

4.2.2. Homo psychologicusH. psychologicus shows a transition from a

fishing society to a mining society. In the firsttime-steps some consumats may have bad luckand harvest such a small number of fish that theybecome dissatisfied. Because this will cause themto engage in deliberation, they perceive mining asa satisfactory alternative opportunity. At aboutt=18 more consumats become dissatisfied andstart mining. The more consumats engage in min-ing, the larger the chances for uncertain con-sumats who are fishing and will also start miningon the basis of imitation or social comparison. Atthat moment the transition from a fishing societyto a mining society becomes manifest. Occasion-ally a mining consumat may engage in fishing fora few time steps on the basis of imitation. How-ever, this will usually lead to dissatisfaction withina few time-steps, and the consumat will go miningagain. Nevertheless, due to imitation, in somesimulation runs a few consumats always remainfishing. At the micro-level a lot of behaviouralchange occurs in the working behaviour of H.psychologicus. At the macro-level a decrease can

be observed in the Gini index from 1 at the startto about 0.25 at t=20. After the transition to amining-society, the Gini index rises again to alevel of about 0.8, indicating that income differ-ences between consumats have decreased as aconsequence of them all engaging in mining.

4.2.3. Homo economicusThe oscillating pattern demonstrated by H. eco-

nomicus (cf. Fig. 7) is caused by the fact that allsixteen consumats change their behaviour at thesame time. This repetitive switching between moremining/less fishing and more fishing/less miningmay be explained by the fact that each consumat,being in the deliberation mode, has been for-malised so as to expect that all others will behaveas in the previous time-step. The consumat thusdeliberates that when all others increase their timespent mining, increasing the time spent fishingmay yield the highest outcomes, and vice-versa.However, all the consumats reason in this direc-tion at the same moment, and as a consequence,they all end up doing about the same in the sametime-step. Over time the extremes in these oscilla-tions converge, indicating that in the long run amore stable working time distribution is likely toemerge. Because all the consumats behave moreor less the same, the Gini index remains at a levelof about 0.99, indicating that the incomes of theconsumats are almost equal.

4.3. Introducing di6ersity in consumats’ abilities

In the previous experimental conditions all con-sumats in both consumat-type conditions hadequal abilities regarding fishing (6600 fish units

Table 1The design of 16 consumats with different fishing and mining abilities

Mining ability

11 13367 89

Consumat 3Consumat 2 Consumat 4Consumat 1Fishing ability 4400Consumat 7 Consumat 85867 Consumat 5 Consumat 6Consumat 11 Consumat 127333 Consumat 9 Consumat 10

Consumat 14 Consumat 158800 Consumat 16Consumat 13


Fig. 11. Time spent fishing and mining for the two-consumattype conditions with consumats having unequal abilities. Fig. 13. The pollution concentration for the two

consumat-type conditions for t=1 to t=50.

per time-step) and mining (100 units of gold pertime-step). However, in real life a differentiationin abilities can easily be observed. The multi-agentapproach makes it very easy to formalise con-sumats having different abilities. The central ques-tion is how such a differentiation affects thebehaviour of the consumats and the ecological–economic system as a whole. To formalise this,the consumats are equipped with differing fishingand mining abilities. Four different levels offishing ability and four different levels of miningability make 4�4=16 different ability settings,thus equipping each consumat with a unique setof abilities. As settings, the average is ratherarbitrarily chosen to be equal to the previousexperimental conditions, the consumats with thehighest ability having twice as much ability as theconsumat with the lowest ability. Table 1 showsthe design with respect to the allocation of thetwo abilities to the sixteen consumats.

As in the previous experiments, the averagemining ability is 100 gold units/time-step and theaverage fishing ability is 6600 fish-units/time-step.The consumats will socially compare to or imitateonly the behaviour of approximately similar otherconsumats. If consumats are engaging in socialprocessing, they compare themselves with con-sumats having about the same quantity of foodand money. Regarding food, the consumat con-siders other consumats having 30% more or lessfood as similar to themselves. Regarding money,this percentage is set at 20%. Formalising similar-ity on the basis of changing variables implies thatthe group of consumats with whom a consumatcompares may change continually, depending onits own state and the state of other ‘similar’consumats. Again, two consumat-type conditionsare created to contrast H. economicus and H.psychologicus.

4.3.1. ResultsThe results of this experiment are depicted in

Fig. 11. These results look very similar to those ofthe previous experiment (Fig. 7). However, closerobservation shows that in the case of H. psycho-logicus the transition from a fishing to a miningsociety takes more time, starting earlier and end-ing with the consumats spending more timefishing than in the previous experiment with equalabilities. This quicker start of the transition iscaused by the consumats with a low fishing abilityand a high mining ability, which will be quicklydissatisfied because of their low fish catch, andthen perceive mining as a more attractive oppor-

Fig. 12. The fish stock for the two consumat-type conditionsfor t=1 to t=50.


tunity. In contrast, the consumats with a highfishing ability and a low mining ability are morelikely to continue fishing, causing that at t=50the time spent fishing is about twice as large as inthe previous experiment. As a consequence, theaverage fish harvest of a population of consumatshaving unequal abilities is larger than the harvestof a population of equal consumats. In otherwords, harvesting behaviour is differentiated de-pending on variation among abilities.

Looking at the consequences for the fish stock(Fig. 12), it can be observed that in the beginningthe fish stock depletes more rapidly in the condi-tion of H. economicus (see further below).

However, despite the fact that H. psychologicusspends less time fishing, after time-step 43 the fishstock depletes more than in the case of H. eco-nomicus. This is caused by the fact that the inten-sive mining by H. psychologicus pollutes the lake,which causes the fish stock to further decrease.

Due to the pollution of the lake (Fig. 13), thefish population decreases, and also the averagecatch decreases. This will make mining a rela-tively more attractive opportunity, and it maythus stimulate consumats to go mining. Evenwhen the pollution causes the fish stock to col-lapse, consumats can satisfy their needs by im-porting fish. It can easily be imagined that insuch a case the eventual depletion of the goldmine will leave the consumats with no opportu-nity at all to satisfy their needs.

Looking at H. economicus, two major differ-ences with the equal abilities experiment (Fig. 7)can be seen. First, the oscillating pattern dissi-pates more quickly in the unequal abilities experi-ment (Fig. 11). This is due to the fact that the

consumats do not behave equally because they allhave different abilities. The resulting behaviouraldiversity means that each consumat has less ex-treme expectations regarding the average timespent working of the other consumats. This in itsturn causes the consumats to behave less ex-tremely for themselves. As a consequence, theoscillations in time spent working converge,thereby further stabilising the expectations, andhence the time spent working. The second differ-ence with the equal abilities experiment is that thetimes spent fishing and mining are lower. Due tothis effect the fish stock is depleted less severelyat time-step 50 than in the condition with equalabilities. The explanation for this is that con-sumats with a high ability do not want to harvestas much as they are capable of, and the con-sumats with a low ability are not capable ofharvesting as much as they want.

In contrast with H. psychologicus, for H. eco-nomicus the average harvest of a population ofconsumats having equal abilities is larger than theharvest of a population of unequal consumats.These results demonstrate that an increase in thediversity of consumat abilities may have differenteffects on the sustainability of a system, depend-ing on the behavioural processing the consumatsengage in. Apparently the effects of increasingdiversity are not unidirectional, which emphasisesthe need for more research to investigate the roleof diversity in resource management situations.

The introduction of diversity has an impact onthe equality of the income of consumats. In Fig.14 the Gini index is presented for H. economicusand the H. psychologicus, both for equal abilitiesand for unequal abilities. It comes as no surprisethat adding variance in the consumat abilitiescauses the Gini index in general to be lower thanin the no-variance condition. Logically, morevariance in ability results in more variance inincome. Moreover, it can be observed that H.psychologicus generally displays a lower Gini in-dex than H. economicus, indicating that non-out-come-optimising behavioural processing strategiescontribute to diversity in behaviour and in result-ing income.

For H. psychologicus with equal abilities therelatively quick transition from fishing to miningFig. 14. The Gini coefficient for equal and unequal abilities.


Fig. 15. The fish stock under conditions of small and largetechnological innovations.

innovation, e.g., engines with more power or bet-ter fishing nets. Once a consumat adopts a newtechnology, the other consumats can also adoptthe new technology. The technology level in-creases slowly towards the highest available tech-nology level among the consumats. In thiscondition all consumats start with equal abilities.Regarding fishing, they start with an ability tocatch 6600 fish-units per capita. Two types ofinnovations have been experimentally tested,namely, small technological innovations, implyingan increase of 100 fish-units in fish catch percapita, and large technological innovations, im-plying an increase of 200 fish-units in fish catchper capita. Again two consumat-type conditionsare created to contrast H. economicus and H.psychologicus.

4.4.1. ResultsFig. 15 shows the development of the fish stock

for H. psychologicus and H. economicus given thetwo types of innovation.

In case of the H. psychologicus, it can be ob-served that the fish stock starts decreasing rela-tively late. This is due to the fact that theconsumats (easily satisfied) do not engage in de-liberation that often, and consequently have asmaller chance of detecting an innovation. As aconsequence, they remain unaware of the innova-tion, causing that the dissipation of the innova-tion through the population of consumatsproceeds at a slower rate. In the case of H.economicus and large technological innovations,in the fish stock collapses very fast because of thefast dissipation of new technology. If a singleconsumat discovers a new innovation, the nexttime-step all other consumats will also be awareof this innovation. In the case of small technolog-ical innovations, the collapse occurs somewhatlater.

These experiments indicate that technical inno-vation that increases the productivity by whichresources are exploited may significantly affect thedepletion rate of a resource. It also suggests thatthe rate at which this happens is strongly tied tothe social conditions under which innovationsmay be diffused.

causes a relatively short period where the oneconsumat is fishing, whereas the other consumatswitches to mining. This yields large income dif-ferences concentrated in a short period of time,which is being reflected in the sharp drop in theGini index around time-step 20. In the conditionof h. psychologicus with unequal abilities, the di-versity in income is larger on the average, as canbe seen from Fig. 14. However, because the tran-sition is not as complete as in the equal-abilitiescondition, and less compressed in time, there willbe no short period of time where the diversity ofincome is being concentrated. Apparently, a fastertransition leads to larger income differencesamong consumats.

4.4. Introducing learning and technologicalinno6ation

The following experiment introduces technolog-ical innovation and learning by consumats toincrease their harvest of fish. Here, mining is notallowed, making the consumats again fully depen-dent on the fish stock for the satisfaction of theirsubsistence, identity and freedom needs. It is as-sumed that technological innovations only emergewhen a consumat is deliberating. During delib-eration there is a random chance that the con-sumat detects a new fishing opportunity thatyields more fish per hour fishing. This is beingformalised as an increase in fishing ability. Thisincrease in fishing ability resembles a technical


5. Conclusions

In mainstream economies, the behaviour ofman in relation to renewable resources is tradi-tionally formalised following the ‘rational actor’approach. In this approach, usually large aggre-gates of people are represented by a single, ratio-nal actor with perfect foresight and a single,individual set of preferences. In this paper weintroduce the consumat approach as an alterna-tive approach, which differs on two aspects fromthe rational actor approach. First, instead of us-ing a single type of decision making, deliberation,as in the rational actor approach, the consumatapproach comprises four cognitive processes,namely deliberation, social comparison, repetitionand imitation. Second, because the formalisationof social processes necessitates the modelling ofseveral interacting agents, a multi agent simulationapproach is being used instead of using a single‘meta-actor’.

Using the consumat approach, we are capableof studying the interactions between a multitudeof agents and their environment. For example, inthis paper we demonstrated that the formalisationof the four cognitive processes in the homo psy-chologicus yields a very different transition from afishing society to a mining society than in case ofthe homo economicus, which could engage in de-liberation only. Moreover, these different transi-tions have consequences for the population of fishin the lake due to overharvesting and/or pollu-tion. The multi-agent structure of the consumatapproach also allows for studying the effects ofdiversity amongst agents (e.g., abilities) on theirinteractions with the environment.

Also, the multi-agent structure allows for thecalculation of quasi macro-variables such as in-come distribution. This provides an interestingheuristic to explore hypotheses between(in)equality and (economic) growth. In successiveexperiments it would be of interest to follow therecommendation by Kumar et al. (1996) of study-ing the volume of poor people, the severity of thepoverty and the distribution of poverty in combi-nation. This would draw a more profound per-spective on distributional equity and thebehavioural dynamics affecting it. Moreover, it

would be interesting to equip the consumats withcultural perspectives (e.g., Thompson et al. 1990;Janssen and De Vries, 1998) or basic orientors(Bossel, 1996; Krebs and Bossel, 1997) and ex-plore how these co-determine the dynamics ofecological–economic systems.

For further application in ecological economicmodels, we suggest that the consumat approachcan be used as a tool box, providing the modellerwith heuristics for introducing relatively simplebehavioural dynamics into integrated models. Inthis way, the consumat approach can be used tosimulate social processes and habitual behaviour,next to deliberate behaviour, in order to unravelthe behavioural dynamics underlying consump-tion of common properties and to design suitablemanagement strategies for our common good.

References

Ajzen, I., 1991. The theory of planned behavior. Organiza-tional Behavior and Human Decision Processes 50, 179–211.

Ajzen, I., Madden, T.J., 1985. Prediction of goal-directedbehaviour: attitudes, intentions, and perceived behavioralcontrol. Journal of Experimental Social Psychology 22,453–474.

Axelrod, R., 1980a. Effective choice in the Prisoner’sDilemma. Journal of Conflict Resolution 24, 3–25.

Axelrod, R., 1980b. More effective choice in the Prisoner’sDilemma. Journal of Conflict Resolution 24, 379–403.

Axelrod, R., 1984. The Evolution of Cooperation. BasicBooks, New York.

Axelrod, R., 1987. The evolution of strategies in the iteratedprisoner’s dilemma. In: Davis, L. (Ed.), Genetic Al-gorithms and Simulated Annealing. Morgan Kaufman,Los Altos, CA, pp. 32–41.

Bandura, A., 1977. Social Learning Theory. Prentice Hall,Englewood Cliffs, NJ.

Bandura, A., 1986. Social Foundations of Thought and Ac-tion: A Social Cognitive Theory. Prentice Hall, EnglewoodCliffs, NJ.

Bossel, H., 1996. 20/20 vision: explorations of sustainablefutures. Center for Environmental Systems Research, Uni-versity of Kassel, Germany.

Brechner, K.C., 1977. An experimental analysis of social traps.Journal of Experimental Social Psychology 13, 552–564.

Brewer, M.B., 1979. In-group bias in the minimal intergroupsituation: a cognitive motivational analysis. PsychologicalBulletin 86, 307–324.

Budescu, D.V., Rapoport, A., Suleiman, R., 1990. Resourcedilemmas with environmental uncertainty and asymmetric


players. European Journal of Social Psychology 20, 475–487.

Cialdini, R.B., Kallgren, C.A., Reno, R.R., 1991. A focustheory of normative conduct: a theoretical refinement andreevaluation of the role of norms in human behavior. In:Berkowitz, L. (Ed.), Advances in Experimental Social Psy-chology, vol 24, pp. 201–234.

Clark, C.W., 1976. Mathematical Bioeconomics: The OptimalManagement of Renewable Resources. Wiley, New York.

Conte, R., Hegselmann, R., Tema, P. (Eds.), 1997. SimulatingSocial Phenomena. Springer, Berlin.

Dasgupta, P.S., Heal, G.M., 1979. Economic Theory andExhaustible Resources. Welwyn (UK) Nisbet and Co.,Cambridge University Press.

De Greef, J., De Vries, H.J.M., 1991. Duurzame ontwikkelingals richtinggevend kader voor milieubeleid, Bilthoven (NL):RIVM, rapportnr. 481501001.

Doran, J., Gilbert, N., 1994. Simulating societies: an introduc-tion. In: Gilbert, N., Doran, J. (Eds.), 1994. SimulatingSocieties: The Computer Simulation of Social Phenomena.UCL Press, London, pp. 1–18.

Edney, J.J., 1980. The commons problem: alternative perspec-tives. American Psychologist 35, 131–150.

Festinger, L., 1954. A theory of social comparision processes.Human Relations 7, 117–140.

Fishbein, M., Ajzen, I., 1975. Belief, Attitude, Intention, andBehaviour: an Introduction to Theory and Research. Ad-dison-Wesley, Reading, MA.

Fox, J., Guyer, M., 1977. Group size and others’ strategy in anN-person game. Journal of Conflict Resolution 21, 323–338.

Gilbert, N., Conte, R., 1995. Artificial Societies: the ComputerSimulation of Social Life. UCL Press, London.

Gilbert, G.N., Troitzsch, K.G., 1999. Simulation for the SocialScientist. Open University Press, Milton Keynes, p. 100.

Gini, C., 1912. VaTiabilita e Mutabilita. Bologna (It.).Guldberg, C.M., Waage, P., 1867. Etudes sur les affinites

chimiques. Christiania.Hardin, R.G., 1968. The tragedy of the commons. Science 162,

1243–1248.Hine, D.W., Gifford, R., 1996. Individual restraint and group

efficiency in commons dilemmas: the effects of two types ofenvironmental uncertainty. Journal of Applied Social Psy-chology 26, 993–1009.

Hogarth, R.M., 1987. Judgement and Choice: the Psychologyof Decision, 2nd. Wiley, New York.

Jager, W., 2000. Modelling consumer behaviour. PhD thesis.Groningen: University of Groningen, Centre for Environ-mental and Traffic Psychology.

Jager, W., Van Asselt, M.B.A., Rotmans, J., Vlek, C.A.J.,Costerman Boodt, P., 1997. Consumer behaviour: a mod-elling perspective in the context of integrated assessment ofglobal change, Globo Report Series No. 17, RIVM,Bilthoven, the Netherlands.

Jager, W., Janssen, M.A., Vlek, C.A.J., 1999. Consumats in acommons dilemma: testing the behavioural rules of simu-lated consumers. COV report no. 99-01. University of

Groningen: Center for Environmental and TrafficPsychology.

Janis, I.L., Mann, L., 1977. Decision Making; A PsychologicalStudy of Conflict, Choice and Commitment. The Free Press,New York.

Janssen, M.A., De Vries, H.J.M., 1998. The battle of perspec-tives: a multi-agent model with adaptive responses to cli-mate change. Ecological Economics 26 (1), 43–65.

Janssen, M.A., De Vries, H.J.M., 1999. Global modelling:managing uncertainty, complexity and incomplete informa-tion. In: van Dijkum, C., de Tombe, D., van Kuijk, E.(Eds.), Validation of Simulation Models. SISWO, Amster-dam, pp. 25–69.

Jerdee, T.H., Rosen, B., 1974. Effects of opportunity to com-municate and visibility of individual decisions on behaviorin the common interest. Journal of Applied Psychology 59,712–716.

Jorgenson, D.O., Papciak, A.S., 1981. The effects of communi-cation, resource feedback, and identifiability on behavior ina simulated commons. Journal of Experimental SocialPsychology 17, 373–385.

Krebs, F., Bossel, H., 1997. Emergent value orientation inself-organisation of an animat. Ecological Modelling 96,143–164.

Kumar, T.K., Gore, A.P., Sitaramam, A.P.G., 1996. Someconceptual and statistical issues on measurements ofpoverty. Journal of Statistical Planning and Inference 49,53–71.

Liebrand, W.P.G., Messick, D.M., (1996), Frontiers in SocialDilemmas Research, Springer Verlag, Berlin.

Liebrand, W.B.G., Nowak, A., Hegselmann, R. (Eds.), 1998.Computer Modeling of Social Processes. Sage Publications,London.

Lindgren, K., Nordahl, M.G., 1995. Cooperation and Commu-nity Structure in Artificial Ecosystems. In: Langton, C.G.(Ed.), Artificial Life, An Overview. The MIT Press, London.

Lorenz, M.O., 1905. Methods of measuring concentration ofwealth. Journal of American Statistical Association 9.

Luce, R.D., Raiffa, H., 1957. Games and Decisions; Introduc-tion and Critical Survey. Wiley, New York.

Machiavelli, N., 1525. Istorie Fiorentine (The Florentine Histo-ries). A series of eight books commissioned by the Floren-tine officials, presented to pope Clement VII.

Macy, M., 1996. Natural selection and social learning inprisoner’s dilemma: co-adaptation with genetic algorithmsand artificial neural networks. In: Liebrand, W.B.G., Mes-sick, D.M. (Eds.), Frontiers in Social Dilemmas Research.Springer Verlag, Berlin.

Malthus, T., 1789. An essay on the principle of population, asit affects the future improvement of society: with remarkson the speculations of Mr. Godwin, M. Condorcet andother writers. London: Johnson.

Masters, J.C., Smith, W.P. (Eds.), 1987. Social Comparison,Social Justice and Relative Deprivation: Theoretical andPolicy Perspectives, vol. 4. Erlbaum, Hillsdale, N.J.

Max-Neef, M., 1992. Development and human needs. In:Ekins, P., Max-Neef, M. (Eds.), Real-life Economics: Un-derstanding Wealth Creation. Routledge, London.


McClintock, C.G., 1978. Social values: Their definition, mea-surement, and development. Journal of Research and De-velopment in Education 12, 121–137.

Messick, D.M., McClintock, C.G., 1968. Motivational basis ofchoice in experimental games. Journal of ExperimentalSocial Psychology 4, 1–25.

Messick, D.M., Allison, S.T., Samuelson, C.D., 1988. Framingand communication effects on group members responses toenvironmental and social uncertainty. In: Maltal, S. (Ed.),Applied Behavioural Economics, vol. 2. Wheatsheaf,Brighton, UK.

Mill, J.S., 1848. Principles of Political Economy: With Some oftheir Applications to Social Philosophy. Reissue (1965).introduced by VW. Bladen; textual editor J.M. Robson,Toronto: University of Toronto Press.

Molander, P., 1985. The optimal level of generosity in aselfish, uncertain environment. Journal of Conflict Resolu-tion 29, 611–618.

Moxnes, E., 1998. Not only the tragedy of the commons:misperceptions of bioeconomics. Management Science 44(9), 1234–1248.

Norton, G.A., 1984. Resource Economics. Edward Arnold,London.

Nowak, M.A., Sigmund, K., 1992. Tit for tat in heterogeneouspopulations. Nature 359, 250–253.

Ostrom, E., Burger, J., Field, C.B., Norgaard, R.B., Polican-sky, D., 1999. Revisiting the commons: local lessons,global challenges. Science 284, 278–282.

Pavlov, I.P., 1927. Conditioned Reflexes. Oxford UniversityPress, New York.

Rapoport, A., Budescu, D.V., Suleiman, R., Weg, E., 1992.Social dilemmas with uniformly distributed resources. In:Liebrand, W.B.G., Messick, D., Wilke, H. (Eds.), SocialDilemmas. Theoretical Issues and Research Findings.Pergamon, Oxford, pp. 43–57.

Ricardo, D., 1817, The Principles of Political Economy andTaxation. Reissue, introduction by Donald Winch (1993)London: Dent.

Rosen, R., 1987. Some epistemological issues in physics andbiology. In: Hiley, B.J., David Peat, F. (Eds.), QuantumImplications: Essays in Honour of David Bohm. Routledgeand Kegan Paul, London.

Rosen, R., 1991. Life Itself, Complexity in Ecological SystemsSeries. Columbia University Press, New York.

Rubenstein, F.D., Watzke, G., Doktor, R.H., Dana, J., 1975.The effects of two incentive schemes upon the conservationof a shared resource by five person groups. OrganizationalBehavior and Human Performance 13, 330–338.

Samuelson, C.D., Messick, D.M., Rutte, C.G., Wilke, H.,1984. Individual and structural solutions to resource dilem-mas in two cultures. Journal of Personality and SocialPsychology 47, 94–104.

Simon, H.A., 1976. Administrative Behavior: A Study ofDecision Making Processes in Administrative Organiza-tions. Harper, New York.

Skinner, B.F., 1953. The Behavior of Organisms. Appleton-Century-Crofts, New York.

Suleiman, R., Rapoport, A., 1989. Environmental and socialuncertainty in single-trial resource dilemmas. Acta Psycho-logica 68, 99–112.

Thompson, M., Ellis, R., Wildavsky, A., 1990. Cultural The-ory. Westview Press, Boulder, USA.

Vallacher, R.R., Nowak, A., 1994. The chaos in social psy-chology. In: Vallacher, R.R., Nowak, R., Nowak, A.(Eds.), Dynamical Systems in Social Psychology. AcademicPress, San Diego.

Walker, C.H., Livingstone, D.R. (Eds.), 1992. Persistent pollu-tants in marine ecosystems. Pergamon Press, Oxford.

Wit, A., Wilke, H., 1998. Public good provision under envi-ronmental and social uncertainty. European Journal ofSocial Psychology 28, 249–256.

.

behaviour in commons dilemmas: homo economicus and homo ...€¦ · studying how micro-level...

Documents