Selection for Attraction

Abstract. In this paper we study the evolutionary potential of a reactive attraction mechanism in a population of agents that interact via cooperation games, as exemplified by the noisy iterated prisoner’s dilemma. Attraction makes an agent unconditionally cooperative toward an attractive opponent, hence introducing a parallel relation between agents, independent from the social interaction context itself, i.e. from the game. An additional partner selection mechanism can exploit such a given psychological condition and discover attracted partners, despite the fact that attraction itself is nowhere represented in the agent reasoning mechanism but is modeled as an uncontrollable process. We show how evolution can select attraction at various levels of increasing complexity and how each level triggers the selection of the next one, provided the corresponding mechanisms emerge behaviorally. We also discuss the implications of these experiments for the study of the evolution of complex cognitive capacities and functionalities