The Autumn Institute in Artificial Intelligence (IA2) is organized by the GDR IA (http://www.gdria.fr). It aims to offer a generalist school in artificial intelligence, in order to provide an overview of the different sub-disciplines of artificial intelligence and their main techniques.  The idea is to provide this more transversal culture of artificial intelligence to our colleagues. This is useful to be able to promote cross-thematic approaches, and cross-fertilization of the different themes of artificial intelligence.

The goal of this school is to provide high-level courses in the different techniques used in artificial intelligence. The courses are divided into fundamental and advanced courses.  The core courses cover the main sub-disciplines of artificial intelligence. These fundamental courses change for each edition, so a colleague can participate to several successive schools to complete his knowledge on the domain.  The advanced courses concern a particular theme (coloration) chosen for each edition of the school. This school is a CNRS thematic school.

The theme chosen for this school, which will take place from September 27 to October 1 in Paris, is "artificial intelligence and explicability", echoing concerns about the explicability of recommendations or decisions made by artificial intelligence algorithms. This dimension of explicability is essential for the acceptability and therefore the large-scale deployment of work resulting from artificial intelligence. 

 

Preliminary program / Programme préliminaire

:: Fundamental courses / Cours fondamentaux:
- An historical perspective on XAI / Une perspective historique sur l’IA explicable (Alain Mille, Université Claude Bernard Lyon)
- Hybrid AI models: some opportunities for explainability / Modèles d’IA hybrides: des opportunités pour l’explicabilité (Isabelle Bloch, LIP6, Sorbonne Université)
- (Deep) neural networks: basic principles and explanation issues / Réseaux de neurones (profonds): principes de base et questions d’explicabilité (Elisa Fromont, IRISA, Université de Rennes)
- Knowledge representation and explainability / Représentation des connaissances et explicabilité (Pierre Marquis, CRIL, Univ. d’Artois)
- Explaining the output of classifiers / Expliquer le résultat de classifieurs (Marie-Jeanne Lesot, LIP6, Sorbonne Université)
- Explainable and Deceptive Behavior for Human-AI Interaction (Sarath Sreedharan, with Subbarao Kambhampati, Arizona State University)

:: Advanced courses / Cours avancés:
- Argumentation, natural language, and explanation / Argumentation, langage naturel et explications (Serena Villata, CNRS, I3S, Université Côte d’Azur)
- Challenges of explanation in the context of autonomous driving / Les défis de l’explicabilité dans le contexte de la conduite autonome (Matthieu Cord, LIP6, Sorbonne Université)
- Explainability in the field: some case studies / Explicabilité sur le terrain: un retour sur quelques cas d’études (Soizic Pénicaud, Simon Chignard, Etalab)
- Challenges in explainability: a Q&A session (Cynthia Rudin, Duke University)