Dynamical Organization of Cooperation in Complex Topologies

In this Letter, we study how cooperation is organized in complex topologies by analyzing the evolutionary (replicator) dynamics of the Prisoner's Dilemma, a two-players game with two available strategies, defection and cooperation, whose payoff matrix favors defection. We show that, asymptotically, the population is partitioned into three subsets: individuals that always cooperate ({\em pure cooperators}), always defect ({\em pure defectors}) and those that intermittently change their strategy. In fact the size of the latter set is the biggest for a wide range of the "stimulus to defect" parameter. While in homogeneous random graphs pure cooperators are grouped into several clusters, in heterogeneous scale-free (SF) networks they always form a single cluster containing the most connected individuals (hubs). Our results show that the paths towards cooperation strongly depend on the underlying structure and give further insights into why cooperation in SF networks is favored.