An introduction to Collective Intelligence

An introduction to Collective Intelligence
NASA tech rep NASA-ARC-IC-99-63, (to appear in J. M. Bradshaw, ed, handbook on agent technology) (Postscript - 1.4 MB )

Abstract: This paper introduces the concept of ``COllective INtelligence'' (COIN). A COIN is a large multi-agent system where: There is little to no centralized communication or control. There is a provided world utility function that rates the possible histories of the full system.. In particular we are interested in COINs in which each agent runs a reinforcement learning algorithm. This approach leads to the following design problem: What rewards functions for the individual agents will, when successfully pursued by those agents result in high world utility?