Multi-Agent Interaction Processes: From Oligopoly Theory to DecentralizedArtificial Intelligence

摘要

The paper addresses similarities between the problems of oligopoly theory anddecentralized artificial intelligence (DzAI). The behavior of autonomous interacting agents is an important issue in both of these areas of research. Among the many questions of interest are ways of reaching cooperation by a group of decentralized agents. The authors think that many models and results of oligopoly theory could prove to be useful in DzAI. Especially, the important idea of reshaping goals with knowledge sharing, or incentive communications, has received little attention in AI. On the other hand, oligopoly theory and experimental economics, in particular, can benefit from the computational methods and tools of AI. To demonstrate this they have built a prototype of an experimental market analysis environment. Some potential goal relaxation techniques for goal conflict situations are illustrated.