On explaining Quantitative Bipolar Argumentation Frameworks

Abstract: Quantitative Bipolar Argumentation Frameworks (QBAFs) provide a powerful tool for modeling reasoning in various applications such as recommender systems and fraud detection. However, there is limited work on explaining their numerical reasoning outcomes in a systematic way. In this talk, I will present three novel explanation methods tailored for QBAFs. First, Argument Attribution Explanations (AAEs) quantify how much each argument contributes to a given outcome. Second, Relation Attribution Explanations (RAEs) shift the focus from explaining the influence of arguments to the support and attack relations, offering a more fine-grained view of the reasoning process. Third, Counterfactual Explanations (CEs) identify changes to the base scores of the arguments that would lead to a different but more desired outcome, supporting actionable insights and contestability.

artificial intelligencelogic in computer sciencelogic

Audience: researchers in the topic

Online logic seminar

Series comments: Description: Seminar on all areas of logic