Integrating Models of Trust, Gossip, and Emotion for Artifical Agents

University Of Florida, Gainesville FL

Investigators

Abstract

This award supports the use of boundedly-rational agent-based models to study the emergence of multi-agent "trusted information structures" in a rudimentary societal group setting by incorporating models of trust and emotion in these agents. As these agents exchange only information, the emergence of these trusted structures represent a primitive (but fundamental) form of organizational learning and adaptation. However, the social setting complicates matters, as the adjustment of these structures is also purely informational - through gossip about the other agents, choices to trust other agents or not, and emotional-type responses to particular categories of events. The computational formulation of the model integrates theories from psychology, sociology, and anthropology. This research advances knowledge at the interstices of social and computational sciences, provides insight into the emergent forms of advice networks occurring on the Internet, explores how elementary forms of emotion alter agent (and group) behaviors, and contributes to the linking of agent properties to emergent group and organizational learning and behaviors.