Mean-field approximation for two- and three-person Prisoner’s Dilemmas

We consider the mean-field approximation (MFA) to the system of interacting agents playing two- and three-person Prisoner’s Dilemma games (2-PD and 3-PD). The agents have three available strategies: All-C, All-D, and a third one, which we choose in various ways. Long-time distribution of the strategies is compared for both games. The latter game admits coexistence of larger number of strategies in the long run. In the case of external pressure for cooperation, more pressure is in general necessary for 3-PD than for 2-PD to guarantee cooperation for almost all initial compositions of the strategies. Results for different types of tit-for-tat strategies for 3-PD are discussed. Unlike the 2-PD case, the more forgiving tit-for-tat strategy can survive in the long run in 3-PD.