Social Dilemmas in Computational Ecosystems

Abstract

Computational ecosystems are large distributed systems in which autonomous agents make choices asynchronously based on locally available information which can be uncertain and delayed. They share these characteristics with biological ecosystems, human societies and market economies. We show that, even when designed with a single overall goal in mind as in the case of distributed problem solving, computational ecosystems can face well-known social dilemmas of sustaining cooperative behavior among selfish agents. Specifically, public-goods problems, where a common good is available to all regardless of individual contribution, can arise due to information limitations as well as the commonly recognized incentive conflicts. Some techniques for mitigating the impact of these problems are also presented. 1.