eXtended Multimodal Composite Association Score (xMCAS): A Gender Inclusive Approach to Measurement of Bias in Text-To-Image Diffusion Models


Abhishek Mandal (Dublin City University), Susan Leavy (University College Dublin), Suzanne Little (Dublin City University)
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)Supplementary

Abstract

Text-To-Image Diffusion Models such as DALL-E and Stable Diffusion have become extremely capable of generating images from text prompts. They have also been shown to exhibit stereotypical gender bias. Previous research has identified and measured the prevalence of gender bias in a binary sense. We introduce eXtended Multimodal Composite Association Score (xMCAS): a novel and easy-to-interpret metric capable of detecting and measuring gender bias adopting a more inclusive concept of gender. Our analysis using this metric revealed the presence of stereotypical concepts of non-binary people in both DALL-E 2 and Stable Diffusion.

Citation

@inproceedings{Mandal_2025_BMVC,
author    = {Abhishek Mandal and Susan Leavy and Suzanne Little},
title     = {eXtended Multimodal Composite Association Score (xMCAS): A Gender Inclusive Approach to Measurement of Bias in Text-To-Image Diffusion Models},
booktitle = {36th British Machine Vision Conference 2025, {BMVC} 2025, Sheffield, UK, November 24-27, 2025},
publisher = {BMVA},
year      = {2025},
url       = {https://bmva-archive.org.uk/bmvc/2025/assets/papers/Paper_182/paper.pdf}
}


Copyright © 2025 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection