RT Dissertation/Thesis
T1 Argumentative Conversational Agents for Explainable Artificial Intelligence
A1 Stepin, Ilia
K1 explainable artificial intelligence
K1 counterfactuals
K1 dialogue game
K1 interpretable fuzzy modelling
K1 human evaluation
AB Recent years have witnessed a striking rise of artificialintelligence algorithms that are able to show outstandingperformance. However, such good performance is oftentimesachieved at the expense of explainability. Not only can the lackof algorithmic explainability undermine the user's trust in thealgorithmic output, but it can also cause adverse consequences.In this thesis, we advocate the use of interpretable rule-basedmodels that can serve both as stand-alone applications andproxies for black-box models. More specifically, we design anexplanation generation framework that outputs contrastive,selected, and social explanations for interpretable (decisiontrees and rule-based) classifiers. We show that the resultingexplanations enhance the effectiveness of AI algorithms whilepreserving their transparent structure.
YR 2023
FD 2023
LK http://hdl.handle.net/10347/31084
UL http://hdl.handle.net/10347/31084
LA eng
DS Minerva
RD 28 abr 2026