Learning versus evolution in iterated prisoner's dilemma

In this paper, we explore interactions in a co-evolving population of model-based adaptive agents and fixed non-adaptive agents playing the iterated prisoner's dilemma (IPD). The IPD is much studied in the game theory, machine learning and evolutionary computation communities as a model of emergent cooperation between self-interested individuals. Each field poses the players' task in its own way, making different assumptions about the degree of rationality of the players and their knowledge of the structure of the game, and whether learning takes place at the group (evolutionary) level or at the individual level. In this paper, we report on a simulation study that attempts to bridge these gaps. In our simulations, we find that a kind of equilibrium emerges, with a smaller number of adaptive agents surviving by exploiting a larger number of non-adaptive ones.