Counterfactual Visual Explanation via Causally-Guided Adversarial Steering

TL;DR

This paper introduces CECAS, a causally-guided adversarial framework for generating high-quality counterfactual visual explanations that respect causal relationships and reduce spurious correlations, improving interpretability.

Contribution

The paper proposes a novel causally-guided adversarial method for counterfactual explanations, addressing limitations of previous approaches by incorporating causal reasoning to enhance explanation quality.

Findings

Outperforms state-of-the-art methods on multiple benchmarks.

Achieves better validity, sparsity, proximity, and realism in explanations.

Effectively reduces spurious correlations in counterfactual images.

Abstract

Recent work on counterfactual visual explanations has contributed to making artificial intelligence models more explainable by providing visual perturbation 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 first 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.