Explainable AI

Explainable AI refers to the ability to set out, in understandable terms, the factors that led a model to a given decision. It is not a matter of comfort: Article 22 of the GDPR gives the data subject the right to obtain human intervention and to contest an automated decision producing legal effects, which presupposes being able to justify it. The European artificial intelligence regulation adds transparency and documentation requirements according to the risk level of the system. In practice, explainability is judged on one thing: being able to say why this particular application was refused.