Evolution and the Backward Induction Problem in the Repeated Prisoners’ Dilemma

Abstract: The finitely Repeated Prisoners’ Dilemma (RPD) is a model of a social dilemma where cooperation is particularly hard to come by. The standard way to solve the game is using backward induction, which by a particular logic gradually eliminates all cooperation from the end of a repeated game. This eliminates cooperation from backwards, with the result that the players defect from the start of the game. This thesis includes two population-based evolutionary models with strategies discussed in the backward induction literature. First, we consider strategies that can eliminate cooperation to a varying degree from backwards. Second, we also include strategies that can act and react to cooperative out-of-equilibrium play in the first step of the RPD. For both of the models we show and examine the conditions under which recurrent cooperation can appear in the population.