Martin Neumann (2008)
Homo Socionicus: a Case Study of Simulation Models of Norms
Journal of Artificial Societies and Social Simulation
vol. 11, no. 4 6
For information about citing this article, click here
Received: 11-Dec-2007 Accepted: 17-Jun-2008 Published: 31-Oct-2008
At the beginning, an individual player has the options to defect (e.g. by cheating in an exam) or not defect. This is accompanied by a certain chance of being observed by other players. The defector receives a certain payoff, while all other players are slightly hurt. Yet, if player j observes the defection of player i, player j can decide to punish (or not to punish) player i. In the case of punishment, player i gets a negative payoff. However, player j has to pay an enforcement cost. The choice of the strategies is dependent on two variables: the boldness Bi, which determines the probability that player i will defect, and the vengefulness Vi , which determines the probability that player i will punish defectors.
The agent population consists of 20 players. Their initial strategy is set at random. Each individual gets four opportunities to defect with a randomly determined probability of being observed. Then the reproduction rate of the players is determined: an individual with a score one standard deviation above the average gets two offspring, an individual with average success gets one offspring, and an individual with a score one standard deviation below the average gets no offspring. These steps are repeated for 100 generations. 5 simulation runs are executed.
One run resulted in a high degree of vengefulness and a low degree of boldness. However, two runs resulted in a moderate level of both boldness and vengefulness, while a further two runs arrived at a state of almost no vengefulness combined with a high degree of boldness.
|Table 1: Axelrod's contribution to answering the questions|
|Table 2: Conte and Castelfranchi's contribution to answering the questions|
On the one hand, Axelrod's model, employing a game theoretic mode of problem description, provides a causal explanation for norm spreading. This includes a designation of mechanisms of norm transmission and normative transformation. An investigation of the functional effect of norms is left aside. Even though the model provides a mechanism for the transformation of the agents, this is not identical with norm internalisation, which remains beyond the scope of this account.
On the other hand, Conte and Castelfranchi's model, utilising cognitive agents in the AI tradition, provides a causal explanation of how norms can have a functional effect on the social level. However, the process of norm spreading is left unanalysed. The result can be summarised in the following table:
|Table 3: Comparison of both accounts|
Coleman investigates the effect of interaction structures on the evolution of co-operation in a prisoner's dilemma situation. Only small groups can prevent the exploitation of strangers.
Macy/Sato examine the effect of mobility on the emergence of trust among strangers in a trust game. While agents with low mobility trust only their neighbours, high mobility supports the evolution of trust among strangers.
Vieth investigates the evolution of fair division of a commodity in an ultimatum game. Including the ability to signal emotions leads to a perfectly fair share. If detection of emotions is costly the proposals even exceed fair share.
Bicchieri et al. present a model of a trust game. It demonstrates how a trust and reciprocate norm emerges in interactions among strangers. This is realised by several different conditional strategies.
Savarimuthu et al. study the convergence of different norms in the interactions of two different societies. Both societies play an ultimatum game against each other. Two mechanisms are examined: a normative advisor and a role model agent.
In this model, a co-ordination and a social dilemma game are examined. Agents learn norms in repeated interactions with different agents. This is denoted as social learning to distinguish this interaction type from repeated games with the same player. The whole population converges to a consistent norm.
This is an extension of the author's first model. The paper studies the interaction of different agent populations. The interaction leads to a breakdown of the beneficent effects of norms, which can only be preserved with the introduction of normative reputation and communication among agents.
Saam and Harrer present an extension of Conte and Castelfranci's model. They investigate the influence of social inequality and power relations on the effectiveness of a 'finder-keeper' norm.
Epstein examines the effect of norms on both the social macro- and the individual micro level. On the macro level, the model generates patterns of local conformity and global diversity. At the level of the individual agents, norms have the effect of relieving agents from individual thinking.
Flentge et al. study the emergence and effects of a possession norm by processes of memetic contagion. The norm is beneficent for the society, but has short-term disadvantages for individual agents. Hence, the norm can only be retained in the presence of a sanctioning norm.
Verhagen tries to obtain predictability of social systems while preserving autonomy on the agent level through the introduction of norms. In the model, the degree of norm spreading and internalisation is studied.
Hales extends the Conte and Castelfranchi model by introducing stereotyping agents. Reputation is projected not on individual agents but on whole groups. This works effectively only when stereotyping is based on correct information. Even slight noise causes the norms to breakdown.
Burke et al. investigate the emergence of a spatial distribution of a binary norm. Patterns of local conformity and global diversity are generated by a decision process dependent on local interaction with neighbouring agents.
|Table 4: Tabular comparison|
|Axelrod (GT)||norm dynamics (norms broadly conceived!)||sanctions||social learning; replicator dynamics||dynamical propensities|
|Colemann(GT)||norm dynamics||punishment by defections (memory restrictions for identifying defections as sanctions)||a) group size (acquaintance)|
b) additionally: replicator dynamics1
|Macy and Sato (GT)||norm dynamics||losses by exclusion from interaction||social learning||dynamical propensities|
|Vieth (GT)||norm dynamics||losses by rejection||social learning; replicator dynamics||dynamical propensities|
|Bicchieri et al. (GT)||norm dynamics||sanctions by retaliating super game strategies||strategy evolution; replicator dynamics||conditional strategies|
|Savarimuthu et al.(GT)||norm dynamics;|
|losses by rejection;|
|Advice updating based on collective experience||dynamical propensities|
|Sen and Airiau (GT)||norm dynamics||experience||social learning guiding behaviour convergence||dynamical propensities|
|Conte and Castelfranchi 95(AI)||functional analysis||–/–2||–/–2||conditional strategies|
|Castelfranchi, Paolucci and Conte 98(AI)||functional analysis||updating conditionals (of strategies) through knowledge||updating knowledge by experience (and communication1)||conditional strategies|
|Saam and Harrer (AI)||functional analysis||a) –/–2|
|Epstein (AI)||norm dynamics; functional analysis||observation||social learning||dynamical updating|
|Flentge et al. (AI)||functional analysis||memetic contagion||contact||conditional strategies|
|Verhagen (AI)||norm dynamics||internalisation||communication||decision tree|
|Hales (AI)||functional analysis||updating conditionals (of strategies) through knowledge||updating knowledge by experience (and communication1)||conditional strategies|
|Burke (AI)||norm dynamics||signals||social learning guiding behaviour convergence||dynamical propensities (Threshold)|
|Σ(Σ |si – gi| / n) m||(1)|
Surprisingly, Verhagen claims, that a higher norm spreading indicates a higher variance of behaviour. Conversely, a higher variance of behaviour is defined as a lower degree of norm internalisation.
2Beside the fundamental paradigm shift towards methodological individualism, Parsons has also been criticised for inherent inconsistencies (compare Balog 2000; Haller 1999; Oakes 1980; Warner 1978; Gouldner 1971; Black 1961). This judgement is emphasised also by evidence from attempts to model Parsons' theory (Jacobsen and Bronson 1997).
3This is in remarkable contrast to Parsons original intention: in fact, he criticised 'utilitarian' theories as deterministic. He claimed that the active role of an actor is “reduced to one of the understanding of his situation and forecasting of its future course of development” (Parsons 1937 , p. 64). Thus, the actor is reduced to a situational automaton. Parsons claimed that a 'voluntaristic' theory of action has to include the active choice of the ends of action. However, in explaining these ends, he relied on the pre-existence of social norms and in so doing reduced the individual actor again to the status of an automaton. Normative orientation is identified with conformity with norms. This is in contradiction to his own approach.
4 For this reason, the work of Brian Skyrms, for instance, is not included. Without a doubt, the evolution of the social contract (Skyrms 1996) is a highly relevant question for the foundation of social norms. However, Skyrms' models remain on the level of population dynamics. Intra agent processes are not taken into regard. The results of the models may be true, but the mechanisms cannot be covered by such an approach. Here we will concentrate on models including in some way intra-agent processes.
5A Replication of a simulation model developed by Robert Axelrod undertaken by Galan and Izquierdo (2005) use analytical tools as well as simulation experiments.
6 For an overview of the broad range of moral dynamics compare Hegselmann (2008). A representative sample of normative architectures is examined by Neumann (2008).
7Yet Galan and Izquierdo (2005) prove that the results are not unequivocal.
8The criticism has been levelled that too much attention has been paid to the concept of a an equilibrium (Merton 1957; Gouldner 1971). This shortcoming is closely related to functional explanations. If a deviation exist from the equilibrium, it is assumed that forces also exist that push the social system into a state of equilibrium again.
9Even though in the models in the AI tradition often punishment is possible, punishment does not induce a transformation of the punished agent.
10The hypotheses are first, that a higher degree of autonomy reduces the predictability of the behaviour. Secondly, that a higher leadership value induces a higher predictability of the behaviour. Thirdly, if the personal decision tree equals the initial group decision tree, it is assumed that the predictability will be higher compared to an initial random group decision tree. In fact, a higher autonomy value leads to a higher degree of norm spreading. Surprisingly, a higher autonomy does not lead to a lower norm internalisation—which indicates (according to Verhagen's assumptions) a higher variance of behaviour. Thus, the Hypothesis is rejected for norm internalisation. The same result holds for the effect of the leadership value: A higher leadership value leads to a higher norm spreading but not to a higher degree of norm internalisation. Only the final assumption is verified for both norm spreading and internalisation.
11Presumably, the assumption is that higher norm spreading indicates a higher difference between the agent's self- and the group-model. However, no reference to the self-model is given in the formula.
12 A similar concept can also be found in the model of Saam and Harrer (1999). The authors deploy the notion of institutionalisation of norms in a society as a whole. Institutionalisation of a norm n means that n is saved in the memory of each agent of the society. Thus, a norm is institutionalised, if it is internalised in each agent of the society. While internalisation refers to the micro-level of individual agents, institutionalisation is a concept that operates on the social macro-level. However, in Saam and Harrer's model, institutionalisation is simply switched on (or off) by a so-called redistribution agent. There is no mechanism at work that could explain the transmission of a norm to individual agents.
13The slower convergence time might be due to the fact that in the case of mechanism two the autonomy value is applied twice: for determining the probability that agents ask for advice, and again for the probability that they accept it.
AXELROD, R. (1986) "an evolutionary approach to norms". American Political Science Review 80 (4) 1095 - 1111.
BALOG, A. (2000) "Theorie als 'theoretisches System': Parsons Beitrag zur soziologischen Theorie". In Staubmann, H., Wenzel, H. (Eds.) Talcott Parsons: Zur Aktualität eines Theorieprogramms, Wiesbaden: Westdeutscher Verlag.
BOMAN, M. (1999) "Norms in artificial decision making". Artificial Intelligence and Law 7 (1) 17 - 35.
BINMORE, K. (1998) "Review of the book: The complexity of cooperation: agent-based models of Competition and Collaboration, by Axelrod, R. Princeton, Princeton University Press". Journal of Artificial Societies and Social Simulation 1 (1) https://www.jasss.org/1/1/review1.html.
BLACK, M. (1961) "Some Questions about Parsons' Theories". In Black, M. (Ed.) The Social Theories of Talcott Parsons, New York: The Free Press.
BOELLA, G., van der Torre, L., Verhagen, H. (2007) "Introduction to Normative Multiagent Systems". Dagstuhl Seminar Proceedings 07122.
BOUDON, R. (1981) The Logic of Social Action. London: Routledge.
CASTELFRANCHI, C., Conte, R., Paolucci, M. (1998) "Normative Reputation and the Costs of Compliance". Journal of Artificial Societies and Social Simulation 1 (3) 3 https://www.jasss.org/1/3/3.html.
COLEMAN, J. (1990) Foundations of Social Theory. Harvard: Belknap.
DAHRENDORF, R. (1956) Homo Sociologicus. Ein Versuch zu Geschichte, Bedeutung und Kritik der Kategorie der sozialen Rolle. Opladen: Westdeutscher Verlag.
DAVIES, K., Moore, W. (1945) "Some Principles of Stratification". American Sociological Review 10 (1) 242 - 249.
DEFFUANT, G., Moss, S., Jager, W. (2006) "Dialogues Concerning a (Possibly) new Science". Journal of Artificial Societies and Social Simulation 9 (1) 1 https://www.jasss.org/9/1/1.html.
DEGUCHI, H. (2001) "Mutual Commitment, Norm Formation and indirect Regulation of Agent Society". Graduate School of economics, Kyoto University Working paper No. 53.
DURKHEIM, E. (1970 ) Regeln der Soziologischen Methode. Neuwied: Luchterhand.
DURKHEIM, E. (1903 ) Erziehung, Moral und Gesellschaft. Vorlesungen an der Sorbonne 1902/1903. Neuwied: Luchterhand.
ESSER, H. (1987) "Warum die Routine nicht weiterhilft—Bemerkungen zur Stagnation soziologischer Theoriebildung". In Mueller, N., Stachowiak, H. (Eds.) Problemloesungsoperator Sozialwissenschaft Vol. 1. Stuttgart: Enke
ESSER, H. (1989) "Verfaellt die 'soziologische Methode'?" Soziale Welt 40, 57 - 75.
ESSER, H. (1993) Soziologie: allgemeine Grundlagen. Frankfurt a. M.: Campus.
ESSER, H., (1996) "What is wrong with 'Variable Sociology'?" European Sociological Review 12 (2) 159 - 166.
FAULBAUM, F. (1992) Von der Variablenanalyse zur Evaluation von Handlungs- und Prozesszusammenhaengen. ZUMA Arbeitsberichte, May 1992.
GALAN, M., Izquierdo, L. (2005) "Appearances can be deceiving: Lessons learned Re-Implementing Axelrod's 'Evolutionary Approach to Norms'". Journal of Artificial Societies and Social Simulation 8 (3) 2 https://www.jasss.org/8/3/2.html.
GELLNER, E. (1971) "Holism versus individualism". In Brodbeck, M. (Ed.) Readings in the Philosophy of the Social Sciences. New York: Macmillan.
GOULDNER, A. (1971) The Coming Crisis of Western Sociology. London: Heinemann.
HALES, D. (2002) "Group Reputation supports beneficent norms". Journal of Artificial Societies and Social Simulation 5 (4) 4 https://www.jasss.org/5/4/4.html.
HALLER, M. (1999) Soziologische Theorie im systematisch-kritischen Vergleich. Leverkusen: Leske + Budrich.
HEGSELMANN, R. (2008) "Moral dynamics." In: Encyclopedia of Complexity and System Science. Berlin: Springer
HOMANS, G. (1964) "Bringing man back in". American Sociological Review 29 (5) 809 - 818.
JACOBSEN, C., Bronson, R. (1997) "Computer simulated empirical tests of Social Theory: Lessons from 15 Years' experience". In Conte, R., Hegselmann, R., Terna, P. (Eds.) Simulating Social Phenomena. Heidelberg: Springer.
MACY, M., Willer, R. (2002) "From Factors to Actors: Computational Sociology and Agent-Based Modelling". American Review of Sociology 28, 143 - 166.
MERTON, R. (1957) Social Theory and Social Structure.2nd Edition. Glencoe: The Free Press.
MOSS, S. (2001) "Game Theory: Limitations and an Alternative". Journal of Artificial Societies and Social Simulation 4 (2) 2 https://www.jasss.org/4/2/2.html.
NEUMANN, M. (2008) "A classification of normative architectures". Proceedings of the World Congress on Social Simulation 2008.
OAKES, G. (1980) Die Grenzen kulturwissenschaftlicher Begriffsbildung. Frankfurt a. M.: Surkamp
PARSONS, T. (1968 ) The Structure of Social Action. A Study in Social Theory with Special Reference to a Group of Recent European Writers. New York, London: Free Press.
PARSONS, T., Shils, E.A. (1951) Towards a General Theory of Action. Harvard: Harvard University Press.
RAUB, W., Voss, Th. (1981) Individuelles Handeln und gesellschaftliche Folgen: Das individualistische Programm in den Sozialwissenschaften. Darmstadt: Luchterhand.
SAAM, N., Harrer, A. (1999) "Simulating norms, social inequality, and functional change in Artificial Societies". Journal of Artificial Societies and Social Simulation 2 (1) 2 https://www.jasss.org/2/1/2.html.
SAVARIMUTHU, B., Purvis, M., Cranefield, S., Purvis, M. (2007) "How do norms emerge in Multi-Agent Societies? Mechanism Design". The Information Science Discussion Paper Series.
SHOHAM, Y., Tenneholtz, M. (1992) "On the synthesis of useful social laws for artificial agent societies (preliminary report)". Proceedings of the Tenth AAAI Conference.
SKYRMS, B. (1996) Evolution of the Social Contract. Cambridge: Cambridge University Press.
SKYRMS, B. (2004) The Stage Hunt and the Evolution of Social Structure. Cambridge: Cambridge University Press.
STALLER, A., Petta, P. (2001) "Introducing Emotions into the Computational Study of Social Norms: a First Evaluation". Journal of Artificial Societies and Social Simulation 4 (1) 2 https://www.jasss.org/4/1/2.html.
VERHAGEN, H. (2001) "Simulation of the Learning of Norms". Social Science Computer Review 19 (3) 296 - 306.
WALLIS, K., Poulton, L. (2001) Internalization: The Origins and Construction of Internal Reality. Buckingham: Open University Press.
WARNER, S. (1978) "Towards a Redefinition of Action Theory: Paying the Cognitive Elements its Due". American Journal of Sociology 83 (6) 1317 - 1349.
WRONG, D. (1961) "The oversocialised conception of man". American Sociological Review 26 92) 183 - 193.
Return to Contents of this issue
© Copyright Journal of Artificial Societies and Social Simulation,