Determining Preferences over Extensions: A Cautious Approach to Preference-Based Argumentation Frameworks

Preferences in abstract argumentation frameworks allow to represent the comparative strength of arguments, or preferences between values that arguments promote. In this paper, we reconsider the approach by Amgoud and Vesic, which computes the extensions of a preference-based argumentation framework by aggregating preferences and attacks into a new attack relation in a way that it favors preferred arguments in conflicts, and then simply applying Dung’s semantics to the resulting graph. We argue that this approach is too rigid in some situations, as it discards other sensible (even if less preferred) alternatives. We propose a more cautious approach to preference-based argumentation, which favors preferred arguments in attacks, but also does not discard feasible alternatives. Our semantics returns a set of extensions and a preference relation between them. It generalizes the approach by Amgoud and Vesic, in the sense that the extensions identified by their semantics will be more preferred than other extensions.