Discovering and Mitigating Visual Biases through Keyword Explanation
This addresses the problem of bias mitigation for AI practitioners by providing an explainable method, though it appears incremental as it builds on existing vision-language models for bias interpretation.
The paper tackles the challenge of identifying and mitigating unexplainable visual biases in computer vision models by proposing the Bias-to-Text (B2T) framework, which interprets biases as keywords extracted from captions of mispredicted images and validates them using vision-language similarity, enabling discovery of biases like gender bias in CelebA and contextual bias in ImageNet.
Addressing biases in computer vision models is crucial for real-world AI deployments. However, mitigating visual biases is challenging due to their unexplainable nature, often identified indirectly through visualization or sample statistics, which necessitates additional human supervision for interpretation. To tackle this issue, we propose the Bias-to-Text (B2T) framework, which interprets visual biases as keywords. Specifically, we extract common keywords from the captions of mispredicted images to identify potential biases in the model. We then validate these keywords by measuring their similarity to the mispredicted images using a vision-language scoring model. The keyword explanation form of visual bias offers several advantages, such as a clear group naming for bias discovery and a natural extension for debiasing using these group names. Our experiments demonstrate that B2T can identify known biases, such as gender bias in CelebA, background bias in Waterbirds, and distribution shifts in ImageNet-R/C. Additionally, B2T uncovers novel biases in larger datasets, such as Dollar Street and ImageNet. For example, we discovered a contextual bias between "bee" and "flower" in ImageNet. We also highlight various applications of B2T keywords, including debiased training, CLIP prompting, and model comparison.