Evolution of Cooperation

Among “Boundedly Rational” Artificial Agents

Uzay Çetin - Boğaziçi University

Department of Computer Engineering

ABSTRACT  We propose to reformulate the payoff matrix structure of Prisoner’s Dilemma Game, by introducing threat and greed factors, and show their effect on the co-evolution of memory and cooperation. We take rather an exploratory approach and use primarily agent-based models to study our complex adaptive system. Our findings are as follows. (i) Memory protects cooperation. (ii) To our surprise, greater memory size is unfavorable to evolutionary success when there is no threat. In the absence of threat, subsequent generations lose their memory and are consequently invaded by defectors. (iii) In contrast, the presence of an appropriate level of threat triggers the emergence of a self-protection mechanism for cooperation, which manifests itself as an increase in memory size within subsequent generations. On the evolutionary level, memory size acts like an immune response of the generations against aggressive defection. This is an un-programmed functionality that emerged via evolutionary dynamics. (iv) Even more extreme threat results again in defection. Our findings boil down to the following: The dose of the threat makes the resistance for cooperation.

BIOGRAPHY  Uzay Çetin received his BSc degree in computer engineering at Galatasaray University. He is selected eligible for French Government scholarship to continue his master on “Artificial Intelligence and Decision" at Pierre Marie Curie University (Paris VI). He completed his PhD in computer engineering department of Boğaziçi University. He is interested in modeling complex adaptive systems, computational social science and artificial intelligence.