Robust Explainable AI: the Case of Counterfactual Explanations

Counterfactual explanations (CXs) are routinely used to shed light on the decisions of machine learning models; however, CX generation strategies often lack robustness, which may jeopardise their explanatory function. In this talk we will begin by introducing the problem of (lack of) robustness and discuss its implications. We will then present some recent solutions we developed to compute CXs with formal robustness guarantees using tools from discrete optimisation and formal verification.