International Workshop on
Regulated Agent-Based Social Systems:
Theories and Applications (RASTA'02)
Abstract for
- 16:30 - 17:30 (second presentation)
- Social Organization in a Software Agent Communtiy with Non-Zero-Sum Game Interaction Model
Matti A. Vanninen and John R. Rose
This paper describes an artificial society of learning software agents employing preferential partner selection and the Iterated Prisoner's Dilemma as a means of social interaction. The IPD has been used extensively in modeling interaction and the development of cooperation in communities, but existing research has focused predominantly on identifying successful strategies in tournaments and ecological simulation. This simulation models an alternative scenario, one in which the members of the community are fixed, but are endowed with a personality and the ability to develop preferences among the other members of the community. It was predicted that such agents would learn to associate with agents which they judge to be favorable, and that the community would achieve a higher net efficiency than it would if interactions occurred randomly. Communities of highly selective and highly competitive agents were found to perform consistently better than non-selective ones.
Daniel Moldt
Last modified: Fri Jun 21 15:53:15 CEST 2002