A Generalized-Bayes Perspective on Counterfactual Explanations: Posterior-Based Decision-Making and Evaluation

2026-07-31

AI Summary

This work addresses the lack of a Bayesian foundation in traditional counterfactual explanations, particularly concerning distance minimization and uncertainty quantification. It establishes, for the first time, a formal connection between counterfactual generation and the generalized Bayesian framework by interpreting it as a maximum a posteriori estimate under a Gibbs posterior with a distance-based prior, leading to the proposed DP-GBCE unified framework. The approach innovatively incorporates risk-averse decision rules—such as Conditional Value-at-Risk (CVaR)—and a multi-model posterior mixture mechanism to balance individual explanation quality with global distributional robustness. Experiments on synthetic and Google Trends data demonstrate trade-offs among different decision rules and introduce comprehensive evaluation metrics that jointly assess both individual counterfactuals and their posterior distributions.

Abstract

Counterfactual explanations (CEs) enhance the interpretability of machine learning models by identifying the smallest change to an input required to obtain a desired output. Although CEs are conventionally formulated as a distance-minimization problem, the theoretical basis of this formulation has received limited attention. We show that a distance-minimization-based CE is mathematically equivalent to the maximum a posteriori (MAP) estimate of a Gibbs posterior within the generalized Bayes framework, specifically when a distance-based prior is used. We call this formulation the Distance-Prior Generalized Bayes CE (DP-GBCE). Building on this posterior perspective, we introduce two decision rules beyond MAP within a unified framework: a Bayes decision that minimizes expected decision loss and CVaR-CE, a risk-averse decision rule. We also propose an extension that uses Bayesian model weights to mix the posterior distributions of multiple models, thereby accounting for model multiplicity, where several models have comparable predictive performance. Finally, we define metrics for evaluating both individual CEs and the posterior distribution as a whole, and use experiments on simulated data and Google Trends data to quantify the trade-offs among the decision rules.

Problem

• Counterfactual Explanations
• Generalized Bayes
• Model Multiplicity
• Posterior Evaluation
• Interpretability

Innovation

• Generalized Bayes
• Counterfactual Explanations
• Posterior Decision Rules
• Model Multiplicity
• CVaR-CE