Pareto Optimality in Abstract Argumentation

Since its introduction in the mid-nineties, Dung’s theory of abstract argumentation frameworks has been influential in artificial intelligence. Dung viewed arguments as abstract entities with a binary defeat relation among them. This enabled extensive analysis of different (semantic) argument acceptance criteria. However, little attention was given to comparing such criteria in relation to the preferences of selfinterested agents who may have conflicting preferences over the final status of arguments. In this paper, we define a number of agent preference relations over argumentation outcomes. We then analyse different argument evaluation rules taking into account the preferences of individual agents. Our framework and results inform the mediator (e.g. judge) to decide which argument evaluation rule (i.e. semantics) to use given the type of agent population involved.