Evolution of strategies in the three-person iterated prisoner's dilemma game.

A three-person iterated prisoner's dilemma (3p-IPD) game is studied. The present simulation uses a lattice model of finite state automata. In particular, the similarity between a noisy 2p- and a noiseless 3p-IPD game is discussed. It is known that, due to the noise, Tit for Tat loses its robustness and is taken over by more complex strategies in a noiseless IPD game. But in the 3p-IPD game, even without noise, Tit for Tat loses its robustness and is also taken over by more complex strategies. It is found that similar strategies take over Tit for Tat in both situations. We thus remark that the role of noise in the two person game is replaced by the third player in the three-person game. As a result, the strategies diversify in both the noisy 2p- and the quiet 3p- IPD game. It is also found that game strategies in an automaton form can be understood as a combination of defensive and offensive substructures. A recognition of these substructures enables us to study the mechanism of robustness in the strategies of the 3p-IPD game.