Subset Team Games and Cooperation Space: Theory and Application , technical report for the 2009 ARL-USMA Technical Symposium, Atlantic City, NJ.
Abstract: A key question in multi-agent systems is the understanding of cooperation among various entities. In many scenarios, it is assumed that individuals compete because they achieve a better individual result. But human behavior is not always so obviously selfish, and there are many examples of systems that are designed specifically for teamwork.
This paper describes work on subset team games, a new mathematical framework useful for studying team-oriented cooperation. Under this generally applicable framework, the results of simulations may be plotted as points in cooperation space, allowing for a high-level analysis of the merits and deficiencies of various algorithms. We will discuss the theoretical underpinnings as well as several applications, including prisoner’s dilemma, a pursuit-evasion games, and social networks.