Dynamics of reciprocal learning by bi-referential model within multiagent systems

In traditional studies of multiagents, the entire group of multiagents has been regarded as a single learning system. It is, however, difficult for us to realize such a learning mechanism in the real world because each agent is exposed only to local interactions with others. Therefore, each agent must acquire a capability, “sociality”, which finds its own role or niche in the social environment, even if the individual´s learning system is self-closed. In this paper, we emphasize that the emergence of “sociality” seems to depend on the dual capabilities of an agent´s referencing; self-referential and social-referential abilities. In addition, we present a learning model of an agent having such dual referencing capabilities as a bi-referential model, in which each referencing capability is implemented by an evolutional computation method of classifier system. We present simulated results obtained by the proposed bi-referential model and also show the results obtained when the available resources are changed. Finally, we discuss the dynamic characteristics of the behaviors emerging within the society of agents