Determining Preferences Through Argumentation

Arguments concerning what an agent should do cannot be considered in isolation: they occur in the context of debates where arguments attacking and defending each other are advanced. This is recognised by the use of argumentation frameworks which determine the status of an argument by reference to its presence in a coherent position: a subset of the arguments advanced which is collectively able to defend itself against all attackers. Where the position concerns practical reasoning, defence may be made by making a choice justified in terms of the values of an agent. Participants in the debate, however, are typically not neutral in their attitude towards the arguments: there will be arguments they wish to accept and others they wish to reject. In this paper we model how a participant in a debate can develop a position which is coherent both with respect to the attack relations between arguments and any value choices made. We define a framework for representing a set of arguments constituting the debate, and describe how a position including the desired arguments can be developed through a dialogue with an opponent. A key contribution is that the value choices are made as part of the argumentation process, and need not be determined in advance.