A co-evolutionary model combined mixed-strategy and network adaptation by severing disassortative neighbors promotes cooperation in prisoner's dilemma games

A co-evolutionary model of both network and mixed strategy is proposed in this study. The assigned strategy s_i of agent i is defined by a real number ranging from 0 to 1, which probabilistically ordains a subsequent action of either cooperation or defection as the agent's offer. We assume a network dynamic to support or hamper the enhancement of cooperation, where an agent severs a link with the neighbor who has the most disassortative strategy. This means that an agent tends to maintain interactions only with neighbors that resemble the agent. A series of numerical simulations reveal that our “assortative grouping” framework enhances cooperation. Interestingly, when a low network adaptation speed and a certain degree of strategy copy error are presumed, phenomenal network heterogeneity evolves, one that realizes more significant cooperation as compared to error-free cases.