Schedule

Abstract: In this talk I will give an overview of computational models of argumentation and I will discuss how these models could be used for explainable AI. Argumentation has both inferential and dialogical aspects. Argumentation as a form of reasoning makes explicit the reasons for the conclusions that are drawn and how conflicts between reasons are resolved with preferences. Argumentation as a form of dialogue is about protocols for resolving conflicts of opinion by verbal means. Both aspects of argumentation have been computationally modelled, and some of this work has been applied to explainable AI.

Abstract: Explainable AI plays a pivotal role in making AI successful in real-life applications, but achieving this requires addressing more than just technical challenges. This talk delves into the "why, who, what, and how" of AI projects, aiming to bridge the gap between academia and industry to foster collaboration for responsible and impactful AI implementations.

Abstract: The increasing complexity of machine learning models calls for explainable artificial intelligence (XAI) methods that not only clarify model predictions but also align with human understanding. In this talk, we explore innovative approaches to XAI, focusing on two key areas: time-series analysis and semantic continuity. Using real-world examples from genomics, predictive maintenance, and counterfactual explanations, we examine how global sensitivity analysis and optimization techniques enhance interpretability across domains. We also introduce the concept of semantic continuity—how minimal semantic changes in input data should correspond to subtle changes in model outputs—and discuss its implications for evaluating and benchmarking XAI methods. Through a mix of case studies and proof-of-concept experiments, we delve into the challenges and opportunities of developing XAI systems that bridge the gap between technical complexity and meaningful insights. This talk will provoke discussion on how to advance fairness, usability, and trust in XAI.

Abstract: In formal argumentation theory, multiple argumentation-based explanation methods have been formulated based on ideas from social and cognitive science. However, these have not yet been empirically validated. To be able to apply argumentation for explanations we need to test whether argumentation-based explanations can be understood by people, and if they do, then which of the available argumentation-based explanation options people prefer. This study describes and empirically validates two types of relatedness used in argumentation-based explanations, related admissibility, and directly related admissibility. This was done by instructing participants to select arguments from an argumentation framework to explain another argument in this framework. These explanations selected by the participants were compared to argumentation-based explanations that use relatedness. We found that both forms of relatedness are cognitively plausible. This gives insight into how argumentation theory can be applied in the real world to provide explanations.

Abstract: The problem of explaining inconsistency-tolerant reasoning in knowledge bases (KBs) is a prominent topic in Artificial Intelligence (AI). In this talk, I present a generic argumentation-based approach to address these problems. This approach is defined for logics involving reasoning with maximal consistent subsets and shows how any such logic can be translated to argumentation. Then our work provides dialogue models as dialectic-proof procedures to compute and explain a query answer wrt inconsistency-tolerant semantics. This allows us to construct dialectical proof trees as explanations, which are more expressive and intuitive than existing explanation formalisms.