Preferential learning and memory resolve social dilemma

Cooperation is widespread in society, thus how to explain this phenomenon has become one open question. According to empirical experience, preferential learning and memory seem to be two effective ways to this issue, which, however, still needs validation in scientific research. Motivate by this point, we consider one-step memory and preference learning (i.e. learning the strategy of subject performing best, which is tuned by a preferential parameter α) in prisoner's dilemma game. α=0 enables the model going back to control treatment where objects randomly selected. While for α > 0, individuals prefer objects that perform better. Compared with control treatment, we find that increasing preferential parameter α can promote cooperative behavior monotonously. In particular, the larger the value of α the stronger and more compact clusters they can form. Finally, in order to investigate the robustness of this mechanism, we also study the evolution of cooperation in small-world network and random regular network.