Selfish memes and selfless agents-altruism in the swap shop

Abstract

A memetic approach is applied to a resource sharing scenario. Agents are represented as cells on a grid applying simple cultural learning rules which selectively replicate and repel memes from neighbours. The memes represent culturally learned traits but also influence resource sharing behaviours. In the experiments presented, multiple cultural groupings which become altruistic towards "in-group" members emerge via a form of group selection. It is demonstrated that such methods can produce more optimal societies than conventional evolutionary methods in given situations. The memetic approach is inherently distributed and dynamic, offering the possibility of application to areas such as collective robotics and software agent co-ordination.