Provably Robust Bayesian Counterfactual Explanations under Model Changes
This addresses the reliability of interpretable AI explanations for users in dynamic real-world settings, though it is incremental as it builds on existing Bayesian CE methods.
The paper tackles the problem of counterfactual explanations becoming invalid when models are updated, by introducing Probabilistically Safe CEs (PSCE) that provide formal probabilistic guarantees for robustness and safety, resulting in more plausible and discriminative explanations compared to state-of-the-art methods.
Counterfactual explanations (CEs) offer interpretable insights into machine learning predictions by answering ``what if?" questions. However, in real-world settings where models are frequently updated, existing counterfactual explanations can quickly become invalid or unreliable. In this paper, we introduce Probabilistically Safe CEs (PSCE), a method for generating counterfactual explanations that are $δ$-safe, to ensure high predictive confidence, and $ε$-robust to ensure low predictive variance. Based on Bayesian principles, PSCE provides formal probabilistic guarantees for CEs under model changes which are adhered to in what we refer to as the $\langle δ, ε\rangle$-set. Uncertainty-aware constraints are integrated into our optimization framework and we validate our method empirically across diverse datasets. We compare our approach against state-of-the-art Bayesian CE methods, where PSCE produces counterfactual explanations that are not only more plausible and discriminative, but also provably robust under model change.