Multi-Group Proportional Representation for Text-to-Image Models

2025-05-29

AI Summary

This paper addresses fairness disparities in text-to-image (T2I) models—specifically, underrepresentation of intersectional demographic groups (e.g., race × gender)—by proposing the Multi-Group Proportional Representation (MPR) framework. We first formally define MPR as a differentiable fairness metric that supports customizable fairness constraints. Our method integrates CLIP-based semantic alignment with fine-grained attribute classifiers to detect intersectional groups and model representation bias; it then employs gradient-driven fine-tuning to directly optimize MPR within the T2I training objective. Evaluated on mainstream models including Stable Diffusion, MPR reduces average representation bias for minority groups by 42% while preserving generation quality (FID degradation <1.5). Our core contribution is the first optimization-friendly, intersectionality-aware fairness metric and its end-to-end trainable paradigm for T2I models.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Ethics, Bias, and Fairness
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy

Application Category

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

Abstract

Text-to-image (T2I) generative models can create vivid, realistic images from textual descriptions. As these models proliferate, they expose new concerns about their ability to represent diverse demographic groups, propagate stereotypes, and efface minority populations. Despite growing attention to the"safe"and"responsible"design of artificial intelligence (AI), there is no established methodology to systematically measure and control representational harms in image generation. This paper introduces a novel framework to measure the representation of intersectional groups in images generated by T2I models by applying the Multi-Group Proportional Representation (MPR) metric. MPR evaluates the worst-case deviation of representation statistics across given population groups in images produced by a generative model, allowing for flexible and context-specific measurements based on user requirements. We also develop an algorithm to optimize T2I models for this metric. Through experiments, we demonstrate that MPR can effectively measure representation statistics across multiple intersectional groups and, when used as a training objective, can guide models toward a more balanced generation across demographic groups while maintaining generation quality.

Problem

• Measure representational harms in T2I image generation
• Ensure proportional representation of intersectional groups
• Optimize models for balanced demographic generation

Innovation

• Introduces Multi-Group Proportional Representation (MPR) metric
• Optimizes T2I models for balanced demographic representation
• Maintains image quality while improving group representation