Hidden Bias in the Machine: Stereotypes in Text-to-Image Models

TL;DR

This study investigates biases in text-to-image models by analyzing generated images across diverse prompts, revealing significant societal stereotypes related to gender, race, and age, and emphasizing the need for more inclusive AI development.

Contribution

The paper introduces a comprehensive methodology for assessing societal biases in T2I models using diverse prompts and compares outputs with real-world images to highlight disparities.

Findings

Significant gender and racial disparities in generated images

Models often reinforce harmful stereotypes present in societal narratives

Need for inclusive datasets to mitigate biases in generative models

Abstract

Text-to-Image (T2I) models have transformed visual content creation, producing highly realistic images from natural language prompts. However, concerns persist around their potential to replicate and magnify existing societal biases. To investigate these issues, we curated a diverse set of prompts spanning thematic categories such as occupations, traits, actions, ideologies, emotions, family roles, place descriptions, spirituality, and life events. For each of the 160 unique topics, we crafted multiple prompt variations to reflect a wide range of meanings and perspectives. Using Stable Diffusion 1.5 (UNet-based) and Flux-1 (DiT-based) models with original checkpoints, we generated over 16,000 images under consistent settings. Additionally, we collected 8,000 comparison images from Google Image Search. All outputs were filtered to exclude abstract, distorted, or nonsensical results. Our analysis reveals significant disparities in the representation of gender, race, age, somatotype, and other human-centric factors across generated images. These disparities often mirror and reinforce harmful stereotypes embedded in societal narratives. We discuss the implications of these findings and emphasize the need for more inclusive datasets and development practices to foster fairness in generative visual systems.