Emergent Social Rationality In A Peer-To-Peer System

Many peer-to-peer (P2P) applications require that nodes behave altruistically in order to perform tasks collectively. Here we examine a class of simple protocols that aim to self-organize P2P networks into clusters of altruistic nodes that help each other to complete jobs requiring diverse skills. We introduce a variant (called ResourceWorld) of an existing model (called SkillWorld) and compare results obtained in extensive (ten billion interactions) simulation experiments. It was found that for both model variants altruistic behavior was selected when certain cost/benefit constraints were met. Specifically, ResourceWorld selects for altruism only when the collective benefit of an action is at least as high as the individual cost. This gives a minimal method for realizing so-called "social rationality," where nodes select behaviors for the good of the collective even though actions are based on individual greedy utility maximization. Interestingly, the SkillWorld model evidences a kind ofsuperaltruismin which nodes are prepared to cooperate even when the cost is higher than the benefit.