Training on Plausible Counterfactuals Removes Spurious Correlations

TL;DR

Training classifiers on plausible counterfactual explanations labeled with incorrect classes can reduce reliance on spurious correlations, improving robustness and fairness.

Contribution

This work extends the paradigm of learning from adversarial examples to plausible counterfactuals, demonstrating enhanced bias reduction and accuracy.

Findings

Classifiers trained on p-CFEs achieve high in-distribution accuracy.

Training on p-CFEs reduces reliance on spurious correlations.

Learning from p-CFEs is more effective than from adversarial perturbations.

Abstract

Plausible counterfactual explanations (p-CFEs) are perturbations that minimally modify inputs to change classifier decisions while remaining plausible under the data distribution. In this study, we demonstrate that classifiers can be trained on p-CFEs labeled with induced \emph{incorrect} target classes to classify unperturbed inputs with the original labels. While previous studies have shown that such learning is possible with adversarial perturbations, we extend this paradigm to p-CFEs. Interestingly, our experiments reveal that learning from p-CFEs is even more effective: the resulting classifiers achieve not only high in-distribution accuracy but also exhibit significantly reduced bias with respect to spurious correlations.