Abstract argumentation and dialogues between agents

Abstract

A multiagent system (MAS) is made up of multiple interacting autonomous agents. It can be viewed as a society in which each agent performs its activity, cooperating to achieve common goals, or competing for them. Thus, every agent has the ability to do social interactions with other agents establishing dialogues via some kind of agent-communication language, under some communication protocol [6]. Argumentation is suitable to model several kind of dialogues in multi-agents systems. Some authors are actually using defeasible argumentation to model negotiation processes between agents [3, 7]. Our current research activities are related to the use of argumentation in agent’s interaction, such as negotiation among several participants, persuasion, acquisition of knowledge and other forms of social dialogue. Usually, argumentation appears as a mechanism to deal with disagreement between agents, for example when some conflict of interest is present. Argumentation can be used, not only to argue about something, but to know more about other agents: it is enough powerfull to play an important role in general social interaction in multiagents systems. The kind of arguments used in dialogues, and their relationship, depends on the type of dialogue involved. According to [8], dialogues can be classified in negotiation, where there is a conflict of interests, persuasion where there is a conflict of opinion or beliefs, indagation where there is a need for an explanation or proof of some proposition, deliberation or coordination where there is a need to coordinate goals and actions, and one special kind of dialogue called eristic based on personal conflicts. Except the last one, all these dialogues may exist in multi-agents systems as part of social activities among agents. Our aim is to define an abstract argumentation framework to capture the behaviour of these different dialogues, and we present here the main ideas behind this task and the new formal definitions. We are not interested in the logic used to construct arguments, nor the comparison method used. Our formulation completely abstracts from the internal structure of the arguments, considering them as moves made in a dialogue. We also consider multiagent systems as a set of multiple interacting autonomous agents.