Mitigate One, Skew Another? Tackling Intersectional Biases in Text-to-Image Models

2025-05-22

AI Summary

Bias in text-to-image (TTI) models exhibits complex cross-dimensional coupling—e.g., between race, gender, and age—where unidimensional debiasing may inadvertently exacerbate or alleviate bias along other dimensions; yet existing methods lack quantitative modeling and controllable intervention of these interdimensional interactions. To address this, we propose BiasConnect, the first framework to quantitatively model causal dependencies among multiple bias dimensions. Building upon it, we introduce InterMit, a user-configurable, interactive debiasing algorithm that employs counterfactual intervention and target-distribution-guided weight optimization. InterMit operates within a modular, training-agnostic architecture to enable coordinated multi-dimensional fairness optimization. Experiments demonstrate that, compared to baselines, our approach reduces bias scores by 36% (0.33 vs. 0.52), accelerates convergence by 24%, and significantly improves generated image quality.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Ethics, Bias, and Fairness
• Philosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Social Networks and Social Media: Fairness and bias in social network and social media analysis
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

The biases exhibited by text-to-image (TTI) models are often treated as independent, though in reality, they may be deeply interrelated. Addressing bias along one dimension - such as ethnicity or age - can inadvertently affect another, like gender, either mitigating or exacerbating existing disparities. Understanding these interdependencies is crucial for designing fairer generative models, yet measuring such effects quantitatively remains a challenge. To address this, we introduce BiasConnect, a novel tool for analyzing and quantifying bias interactions in TTI models. BiasConnect uses counterfactual interventions along different bias axes to reveal the underlying structure of these interactions and estimates the effect of mitigating one bias axis on another. These estimates show strong correlation (+0.65) with observed post-mitigation outcomes. Building on BiasConnect, we propose InterMit, an intersectional bias mitigation algorithm guided by user-defined target distributions and priority weights. InterMit achieves lower bias (0.33 vs. 0.52) with fewer mitigation steps (2.38 vs. 3.15 average steps), and yields superior image quality compared to traditional techniques. Although our implementation is training-free, InterMit is modular and can be integrated with many existing debiasing approaches for TTI models, making it a flexible and extensible solution.

Problem

• Analyzing interrelated biases in text-to-image models
• Quantifying bias interactions using counterfactual interventions
• Developing an intersectional bias mitigation algorithm

Innovation

• BiasConnect analyzes bias interactions quantitatively
• InterMit reduces bias with fewer steps
• Modular training-free integration with debiasing methods