Gender Artifacts from Art History to Text-to-Image Generation

2026-06-04

AI Summary

This study addresses the prevailing tendency in AI research to treat artistic style as a superficial visual attribute, thereby overlooking its deep entanglement with gender representation within sociohistorical contexts. To bridge this gap, the authors introduce StyleGender—the first dataset integrating art historical knowledge with text-to-image generation, comprising 74,000 images across 19 distinct artistic styles—and propose two novel metrics, PixelSGA and MaskSGA, to quantify gendered visual cues at both pixel and structural levels. Their analysis reveals that artistic style profoundly shapes gender representation: style-related prompts systematically inject historical gender norms into generated imagery, and contemporary generative models often amplify these gender stereotypes more intensely than the original historical artworks themselves.

Abstract

Artistic styles are rooted in specific socio-historical contexts that encode social hierarchies, including distinct constructions of gender. Yet in AI research, style has long been treated as a surface-level visual property: a filter of color, brushstroke, and texture applied to otherwise content-neutral scenes. We introduce the first dataset to investigate the interplay between gender representation and style in both historical and generated images. StyleGender comprises 74k images spanning 19 artistic styles, comprising art historical images with style and gender annotations, T2I-generated images under controlled style and gender prompts, and a semantically aligned set enabling direct art history-to-generation comparison. By proposing two Set Gender Artifact (SGA) metrics (PixelSGA and MaskSGA), capturing gender signals at the pixel level and in compositional structure, we show that (1) gender representation shapes visual features across artistic styles, (2) style keywords carry these patterns into T2I generation, and (3) generative models tend to amplify gender artifacts beyond what is observed in historical sources.

Problem

• gender artifacts
• text-to-image generation
• artistic style
• gender representation
• generative bias

Innovation

• gender artifacts
• text-to-image generation
• artistic style
• StyleGender dataset
• Set Gender Artifact (SGA) metrics