Agents adapt to majority behaviours

Abstract

Agents within a group can have different perceptionsof their working environment and autonomously fulfiltheir goals. However, they can be aware of beliefsand goals of the group as well as other members sothat they can adjust their behavioursaccordingly. To model these agents, we explicitlyinclude knowledge commonly shared by the group andthat obtained from other agents. By avoiding actionswhich violate ``mental attitudes'' shared by themajority of the group, agents demonstrate theirsocial commitment to the group. Defeasible logic ischosen as our representation formalism for itscomputational efficiency, and for its ability tohandle incomplete and conflictinginformation. Hence, our agents can enjoy the lowcomputational cost while performing ``reasoningabout others''. Finally, we present theimplementation of our multi-agent system.