A Large Scale Analysis of Gender Biases in Text-to-Image Generative Models

2025-03-30

AI Summary

This work systematically uncovers gender bias in text-to-image generative models across everyday contexts. To this end, the study constructs 3,217 gender-neutral prompts and generates over 2.29 million images using DALL·E 2/3, Stable Diffusion, Imagen, and Midjourney—enabling the first large-scale, multi-scenario bias assessment across domestic, financial, caregiving, and technical labor domains. Methodologically, it introduces a novel semantic clustering quantification framework based on BERT embeddings and hierarchical clustering, moving beyond traditional binary gender analysis to jointly model objects, activities, and scenes. Results reveal pronounced disparities: women are significantly overrepresented in caregiving and interpersonal contexts (+42%), while men dominate technical and physical labor (+38%); critically, women constitute less than 19% of individuals depicted in finance-related imagery, exposing systemic representational gaps.

Technology Category

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

Application Category

• Social Networks and Social Media: Fairness and bias in social network and social media analysis
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

With the increasing use of image generation technology, understanding its social biases, including gender bias, is essential. This paper presents the first large-scale study on gender bias in text-to-image (T2I) models, focusing on everyday situations. While previous research has examined biases in occupations, we extend this analysis to gender associations in daily activities, objects, and contexts. We create a dataset of 3,217 gender-neutral prompts and generate 200 images per prompt from five leading T2I models. We automatically detect the perceived gender of people in the generated images and filter out images with no person or multiple people of different genders, leaving 2,293,295 images. To enable a broad analysis of gender bias in T2I models, we group prompts into semantically similar concepts and calculate the proportion of male- and female-gendered images for each prompt. Our analysis shows that T2I models reinforce traditional gender roles, reflect common gender stereotypes in household roles, and underrepresent women in financial related activities. Women are predominantly portrayed in care- and human-centered scenarios, and men in technical or physical labor scenarios.

Problem

• Analyzes gender bias in text-to-image generative models
• Examines gender stereotypes in daily activities and contexts
• Assesses underrepresentation of women in financial activities

Innovation

• Large-scale dataset of gender-neutral prompts
• Automatic gender detection in generated images
• Semantic grouping for bias analysis