Supervision : Marie-Jeanne LESOT, Christophe MARSALA

Co-supervision : LAUGEL Thibault (AXA)

Interprétabilité des modèles en apprentissage automatique

This thesis belongs to the field of eXplainable AI (XAI). We focus on post-hoc interpretability methods that aim to explain to a user the prediction made for a specific data instance by a trained decision model.
To increase the interpretability of explanations, this thesis studies the integration of user knowledge into these methods, and thus aims to improve the comprehensibility of the explanation by generating personalized explanations adapted to each user. To achieve this, we propose a general formalism that explicitly integrates knowledge via a new criterion in the interpretability objectives. This formalism is then declined for different types of knowledge and different types of explanations, particularly counterfactual examples, leading to the proposal of several algorithms (KICE, Knowledge Integration in Counterfactual Explanation, rKICE for its variant including knowledge expressed by rules and KISM, Knowledge Integration in Surrogate Models).
The issue of aggregating classical quality and knowledge compatibility constraints is also studied, and we propose to use Gödel's integral as an aggregation operator.
Finally, we discuss the difficulty of generating a single explanation adapted to all types of users and the notion of diversity in explanations.

2022-2024 Publications

• 2024
  • A. Jeyasothy : “Interprétabilité des modèles en apprentissage automatique”, thesis, phd defence 02/20/2024, supervision Lesot, Marie-Jeanne Marsala, Christophe, co-supervision : Laugel, Thibault (AXA) (2024)
• 2023
  • A. Jeyasothy, A. Rico, M.‑J. Lesot, Ch. Marsala, Th. Laugel : “Intégration de connaissances en XAI avec les intégrales de Gödel”, LFA 2023 - RENCONTRES FRANCOPHONES SUR LA LOGIQUE FLOUE ET SES APPLICATIONS, Bourges, France, (Cépaduès) (2023)
  • A. Jeyasothy, Th. Laugel, M.‑J. Lesot, Ch. Marsala, M. Detyniecki : “A General Framework for Personalising Post Hoc Explanations through User Knowledge Integration”, International Journal of Approximate Reasoning, vol. 160, pp. 108944, (Elsevier) (2023)
  • A. Jeyasothy, A. Rico, M.‑J. Lesot, Ch. Marsala, Th. Laugel : “Knowledge Integration in XAI with Gödel Integrals”, International Conference on Fuzzy Systems, Incheon, Korea, Republic of (2023)
  • Th. Laugel, A. Jeyasothy, M.‑J. Lesot, Ch. Marsala, M. Detyniecki : “Achieving Diversity in Counterfactual Explanations: a Review and Discussion”, FAccT ’23, vol. 23 (6), Chicago, Il, United States, (ISBN: 979-8-4007-0192-4) (2023)
• 2022
  • A. Jeyasothy, Th. Laugel, M.‑J. Lesot, Ch. Marsala, M. Detyniecki : “Intégration de connaissances dans les méthodes d’explications post-hoc”, Rencontres francophones sur la logique floue et ses applications, Toulouse, France (2022)
  • A. Jeyasothy, Th. Laugel, M.‑J. Lesot, Ch. Marsala, M. Detyniecki : “Integrating Prior Knowledge in Post-hoc Explanations”, Information Processing and Management of Uncertainty in Knowledge-Based Systems (IPMU'2022), vol. 1602, Communications in Computer and Information Science, Milan, Italy, pp. 707–719, (Springer) (2022)