Biological and Social Symbiosis: An-Agent Based Exploration


    Game theory provides a general and useful framework through which to understand the phenomenon of symbiosis.  Although symbiosis has been approached from a game theoretic perspective - which necessarily focuses on payoffs to individual agents – the evolution of symbiosis has not yet been simulated through an agent-based model.  In nature the payoffs for symbiotic interactions are not externally defined, but arise out of agent properties and interactions.  By using an agent-based model, one can recreate this local agent-to-agent interaction.  
    In this model of symbiosis, game-theoretic payoffs evolve over time.  Changes in these payoffs cause qualitative changes in the games, through changing the incentives for cooperation and defection.  These qualitative changes might correspond to the changing payoffs which accompany the evolution of mutualism from parasitism and vice versa.  In addition to providing a model of the evolution of interspecies mutualism, this simulation could also be interpreted as a model of social interactions among different types of individuals within the same species.  While the evolution of cooperation has been approached extensively from a game theoretic perspective where games are fixed, this simulation extends this work to include the possibility that these games can change qualitatively over time.

C. Athena Aktipis, Martin Zwick, Jeff Fletcher, Wayne Wakeland


Cooperative agents with Contingent Movement Rules Outperform Tit-for-Tat and PAVLOV


    Over evolutionary history, humans and many other animals have evolved cognitive and behavioral mechanisms in order to navigate in social environments.  However, evolutionary psychologists do not yet know exactly which strategies humans and other animals use to avoid interactions with exploitative individuals.  It is also unclear exactly how individuals evolved to be cooperative when cooperation is costly.  Several different strategies have been suggested (tit-for-tat, memory of interactions, reputation effects, subjective commitments) but this is only a small subset of the possible strategies that might have favored the evolution of cooperation.  Before assuming that cooperation evolved in humans because of sophisticated information processing abilities, one should consider the possibility that less complex strategies could have favored the evolution of cooperation.  This series of computer simulations shows that a behavioral rule that is even simpler than Tit-for-Tat and PAVLOV can outperform Tit-for-Tat, PAVLOV and defecting strategies.  In the first experiment, cooperative agents with contingent movement rules (to move after their partner defected) achieved a higher frequency than indiscrimate defectors and defectors that used contingent movement rules to minimize interactions with other defectors.  The second and third experiments showed that this same contingent movement strategy can outperform both Tit-for-Tat and PAVLOV.  This contingent movement strategy was successful, despite its simplicity, because it exploited various aspects of the physical and social environment.  Limitations of the strategy and the experimental model are discussed.


C. Athena Aktipis
Portland State University
Systems Science Department
http://www.sysc.pdx.edu/aktipis/
aktipis@alumni.reed.edu