Abstract: The pervasiveness of AI demands their responsible use with eXplainable AI (XAI). While myriad explanation techniques have been proposed, they remain insufficient to support human understanding due to their poor fit and usability. I present methods to make XAI more human-like by following processes from human psychology. Furthermore, the real-world use of AI brings many requirements which can conflict with explainability. I discuss several confounds that diminish the effectiveness of XAI, such as cognitive load, and privacy, and present methods to mitigate them. Therefore, by studying human reasoning and requirements, human-centric XAI can be made more interpretable and trustworthy.

Abstract: Providing high-quality explanations for AI predictions is a challenging task. It requires, among other elements, selecting a proper level of generality/specificity of the explanation, referring to specific elements that have contributed to the decision, and providing evidence supporting negative hypothesis. In this talk, I will present some results achieved in the area of Argument Mining and Argument Generation and how these results can be exploited – in general and in legal applications – to generate high-quality explanatory dialogues crucially based on argumentation mechanisms.