Subject PostDoc

The postdoctoral researcher will work in the field of trustworthy AI and explainable AI. The successful candidate will focus on implementing concept-based approaches, which aim to identify concepts within neural networks. These approaches are widely used in the context of multimodal data (text/image). The novelty here will be to introduce causal relationships or dependencies in the latent layer between concepts. Other possibilities involve exploring hierarchical relationships between concepts to enhance interpretability. This topic is central to the CAUSALITAI project of the PEPR AI initiative, in collaboration with the Interdisciplinary Laboratory of Digital Sciences at the University of Paris-Saclay. Additional potential collaborations are possible with the ANR ClearDeep project and the MICS laboratory at Centrale Supelec.