Co-argumentation Artifact for Agent Societies

摘要

In a social context, people have only partial knowledge about the world and use arguments in order to solve problems, to reduce conflicts, or to exchange information. Argumentation is a dialogic process, and could occur through direct interaction, or through supports of some sorts-like blackboards, or electronic fora. The same holds for intelligent agents in a multi-agent system (MAS); here, however, it is not clear what could act as a support for argumentation between agents, external to the agents themselves. To this end, this work exploits the agents and artifacts (A&A) meta-model for MAS, exploring the use of artifacts for agent argumentation within a MAS. Along this line, the first aim of this work is to design an argumentation component based on Dung's preferred semantics, combining it with artifact abstraction in order to realise a social support for argumentation in MAS. Using argumentation within the A&A meta-model, we introduce here the notion of Co-Argumentation Artifact (CAA) as an artifact specialised in managing arguments and providing a coordination service for argumentation process in a MAS. In order to give concreteness to our proposal, we also discuss a first CAA deployment based on logic programming and tuple centres exploiting the TuCSoN infrastructure.