Supervision : Isabelle BLOCH

Co-supervision : LESOT Marie-Jeanne

Explicabilité des systèmes d'intelligence artificielle hybride dans un contexte d'interactions avec des humains

This thesis deals with explainability of artificial intelligence systems, in a context of interactions with humans. The objective is to formalize the notion of explanation when both data and knowledge are involved, as it is typically the case in many situations of interactions with humans. A first step will be to structure the space of explanations, by proposing measures to compare explanations. What is a good, or preferred, explanation is another question to be addressed. This will require the definition of a partial ordering relation on the space of explanations. The thesis will focus on conversational agents. Explanations will concern the behavior of the agent, the behavior of the human, and will help improving interactions between them.

2022-2023 Publications