The Digital Mirror: Gender Bias and Occupational Stereotypes in AI-Generated Images
Siiri Leppälampi, Sonja M. Hyrynsalmi, Erno Vanhala
TL;DR
The paper examines gender bias in AI-generated occupational imagery by prompting two diffusion-based generators, DALL·E 3 and Ideogram, across Finnish occupation categories and additional student/professor prompts. It employs Braun & Clarke's thematic analysis to extract patterns, finding that both models reinforce traditional gender stereotypes, with women depicted as younger and more emotive and men as older and more analytical. DALL·E 3 shows stronger gender bias by underrepresenting women in male-dominated and balanced occupations and by underrepresenting older women, while Ideogram more closely matches real-world gender distributions. The study highlights the ongoing risk of bias in AI-generated visuals, advocates for improved training data, fairness controls, and continuous evaluation to foster more representative and inclusive AI visualisations in professional contexts.
Abstract
Generative AI offers vast opportunities for creating visualisations, such as graphics, videos, and images. However, recent studies around AI-generated visualisations have primarily focused on the creation process and image quality, overlooking representational biases. This study addresses this gap by testing representation biases in AI-generated pictures in an occupational setting and evaluating how two AI image generator tools, DALL-E 3 and Ideogram, compare. Additionally, the study discusses topics such as ageing and emotions in AI-generated images. As AI image tools are becoming more widely used, addressing and mitigating harmful gender biases becomes essential to ensure diverse representation in media and professional settings. In this study, over 750 AI-generated images of occupations were prompted. The thematic analysis results revealed that both DALL-E 3 and Ideogram reinforce traditional gender stereotypes in AI-generated images, although to varying degrees. These findings emphasise that AI visualisation tools risk reinforcing narrow representations. In our discussion section, we propose suggestions for practitioners, individuals and researchers to increase representation when generating images with visible genders.