Practical Issues in Detecting Broken Social Commitments

J. Heard and R. Kremer (Canada)


social commitments, conversation policies, CASA, multi agent systems


An open system should admit agents from many sources and these agents may have conflicting goals. Therefore, some actions that an agent would like to perform could be detrimental to other agents. Such actions can be either ac ceptable or unacceptable within a given system. So there should be a way to limit the actions of agents to reduce the occurrence of unacceptable behaviour. One way to begin to accomplish this goal is to have the system observe the actions of agents to model their behaviour. Antisocial be haviours could then be detected, and some action could be taken to prevent agents from performing further unaccept able actions. In this paper we discuss the use of social commit ments to allow a system to detect unacceptable behaviour. Social commitments model an agent’s commitments within a society. Some are implied while others are explicitly stated. Our system defines unacceptable behaviour as any action that breaks an agent’s own social commitments. This paper focusses on the practical requirements that must be met for a system to implement social commitments as a way of detecting unacceptable actions. In addition, we give an overview of how our multi-agent system design supports this goal.

Important Links:

Go Back