Leaving the Prison: A Discussion of the Iterated Prisoner's Dilemma under Preferential Partner Selection

In many situations that can be characterized as a prisoner's dilemma (PD) agents are not prisoners forced to play the game with whoever they face but are free to select their game partners. Opening the prison door adds an interesting dimension to the original game: players face two strategic choices: how to play - their PD strategy - and whom to play with. Those choices are interlinked for whom to play with depends on your PD strategy while the success of this strategy depends on the distribution of strategies in the population - on the potential game partners. This paper examines with the help of simulations how the choice possibility affects the performance of fixed strategies represented as finite automata and the social network emerging from preferential partner selection. As such it serves as a basis for the more compete analysis in which players can also revise their PD strategies. It is shown that when strategy revision is gradual and limited to a certain given set of strategies strategy revision combined with preferential partner selection leads to stable cooperative play. In the repeated context an intermediate case between free and forced interactions is also examined in which players cannot play everyone who is willing to play them nor refuse every unacceptable offer.