Counterfactual Visual Explanation via Causally-Guided Adversarial Steering

Yiran Qiao, Disheng Liu, Yiren Lu, Yu Yin, Mengnan Du, Jing Ma

Keywords: Explainable AI, Counterfacutural Generation, Causality

TL;DR: This paper proposes CECAS, a causally guided framework for counterfactual visual explanations that avoids spurious changes and generates higher-quality counterfactuals than existing methods.

Abstract: Recent work on counterfactual visual explanations has contributed to making artificial intelligence models more explainable by providing visual perturbations to flip the prediction. However, these approaches neglect the causal relationships and the spurious correlations behind the image generation process, which often leads to unintended alterations in the counterfactual images and renders the explanations with limited quality. To address this challenge, we introduce a novel framework CECAS, which leverages a causally-guided adversarial method to generate counterfactual explanations. It innovatively integrates a causal perspective to avoid unwanted perturbations on spurious factors in the counterfactuals. Extensive experiments demonstrate that our method outperforms existing state-of-the-art approaches across multiple benchmark datasets and ultimately achieves a balanced trade-off among various aspects of validity, sparsity, proximity, and realism.

Primary Area: interpretability and explainable AI

Submission Number: 8172