Computing argumentative explanations in bipolar argumentation frameworks

• Main Authors: Miao, Chunyan, Leung, Cyril, Shen, Zhiqi, Chin, Jing Jih, Zeng, Zhiwei
• Other Authors: School of Computer Science and Engineering
• Format: Conference or Workshop Item
• Language: English
• Published: 2019
• Subjects: Argumentation; Explainable Artificial Intelligence; Social sciences::Sociology
• Online Access: https://hdl.handle.net/10356/103315; http://hdl.handle.net/10220/49774
• Institution: Nanyang Technological University

The process of arguing is also the process of justifying and explaining. Transparent reasoning process endows argumentation good explainability. Recently, more research efforts have been devoted to realizing the explanatory power of argumentation in unipolar argumentation frameworks. In addition to the attack relation, bipolar frameworks consider the support relation, which brings greater expressibility but also complexity. It is worth exploring how the interactions encompassed in the support relation contribute to the arguing process and how to capture them in explanations. In this paper, we propose a “stronger” notion of defence and a new bipolar admissibility semantics, which are defined based on both the attack and the support relations, and use them to formalize two types of explanations, namely concise and strong explanations. We then present complete and sound processes for computing explanations by constructing bipolar dispute trees.