Social Learning in Networked Agent Societies

摘要

In networked multiagent societies, how global social order (i.e., social norm) can be achieved through agents' local interactions is a critical research problem in multiagent systems. It has been shown that learning from individual local interactions is an effective mechanism to facilitate norm emergence. Most of the existing work, however, mainly focuses on studying norm emergence via agent learning from its individual experience. The role of social learning, i.e., learning directly from others, has been comparatively less investigated. This paper steps forward the state-of-the-art by investigating how social learning can impact the emergence of social norms in networked agent societies. Experiments are carried out to show the impact of different strategies of choosing between individual and social learning on norm emergence. Experimental results reveal some significant insights into the manipulation and control of norm emergence in networked agent societies achieved through agent local behaviors.