eXplainable Artificial intelligence: From machine to humans, how to make them collaborate?
Résumé
Whether it is upstream, when providing data, when implementing an architecture, or when using algorithms, humans impact AI algorithms through their cognitive biases, their habits and through the data that the algorithms learn. And in turn, the latter impact users when they are used on a large scale. The talk is an introduction to the field of explainable AI with a focus on the impact of biases on AI models and an overview of XAI techniques.