Exploring Social Bias in Downstream Applications of Text-to-Image Foundation Models

Adhithya Prakash Saravanan, Rafal Kocielnik, Roy Jiang, Pengrui Han, Anima Anandkumar
Proceedings on "I Can't Believe It's Not Better: Failure Modes in the Age of Foundation Models" at NeurIPS 2023 Workshops, PMLR 239:84-102, 2023.

Abstract

Text-to-image diffusion models have been adopted into key commercial workflows, such as art generation and image editing. Characterizing the implicit social biases they exhibit, such as gender and racial stereotypes, is a necessary first step in avoiding discriminatory outcomes. While existing studies on social bias focus on image generation, the biases exhibited in alternate applications of diffusion-based foundation models remain under-explored. We propose a framework that uses synthetic images to probe two applications of diffusion models, image editing and classification, for social bias. Using our framework, we uncover meaningful and significant inter-sectional social biases in Stable Diffusion, a state-of-the-art open-source text-to-image model. Our findings caution against the uninformed adoption of text-to-image foundation models for downstream tasks and services.

Cite this Paper

BibTeX
@InProceedings{pmlr-v239-saravanan23a, title = {Exploring Social Bias in Downstream Applications of Text-to-Image Foundation Models}, author = {Saravanan, Adhithya Prakash and Kocielnik, Rafal and Jiang, Roy and Han, Pengrui and Anandkumar, Anima}, booktitle = {Proceedings on "I Can't Believe It's Not Better: Failure Modes in the Age of Foundation Models" at NeurIPS 2023 Workshops}, pages = {84--102}, year = {2023}, editor = {Antorán, Javier and Blaas, Arno and Buchanan, Kelly and Feng, Fan and Fortuin, Vincent and Ghalebikesabi, Sahra and Kriegler, Andreas and Mason, Ian and Rohde, David and Ruiz, Francisco J. R. and Uelwer, Tobias and Xie, Yubin and Yang, Rui}, volume = {239}, series = {Proceedings of Machine Learning Research}, month = {16 Dec}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v239/saravanan23a/saravanan23a.pdf}, url = {https://proceedings.mlr.press/v239/saravanan23a.html}, abstract = {Text-to-image diffusion models have been adopted into key commercial workflows, such as art generation and image editing. Characterizing the implicit social biases they exhibit, such as gender and racial stereotypes, is a necessary first step in avoiding discriminatory outcomes. While existing studies on social bias focus on image generation, the biases exhibited in alternate applications of diffusion-based foundation models remain under-explored. We propose a framework that uses synthetic images to probe two applications of diffusion models, image editing and classification, for social bias. Using our framework, we uncover meaningful and significant inter-sectional social biases in Stable Diffusion, a state-of-the-art open-source text-to-image model. Our findings caution against the uninformed adoption of text-to-image foundation models for downstream tasks and services.} }
Endnote
%0 Conference Paper %T Exploring Social Bias in Downstream Applications of Text-to-Image Foundation Models %A Adhithya Prakash Saravanan %A Rafal Kocielnik %A Roy Jiang %A Pengrui Han %A Anima Anandkumar %B Proceedings on "I Can't Believe It's Not Better: Failure Modes in the Age of Foundation Models" at NeurIPS 2023 Workshops %C Proceedings of Machine Learning Research %D 2023 %E Javier Antorán %E Arno Blaas %E Kelly Buchanan %E Fan Feng %E Vincent Fortuin %E Sahra Ghalebikesabi %E Andreas Kriegler %E Ian Mason %E David Rohde %E Francisco J. R. Ruiz %E Tobias Uelwer %E Yubin Xie %E Rui Yang %F pmlr-v239-saravanan23a %I PMLR %P 84--102 %U https://proceedings.mlr.press/v239/saravanan23a.html %V 239 %X Text-to-image diffusion models have been adopted into key commercial workflows, such as art generation and image editing. Characterizing the implicit social biases they exhibit, such as gender and racial stereotypes, is a necessary first step in avoiding discriminatory outcomes. While existing studies on social bias focus on image generation, the biases exhibited in alternate applications of diffusion-based foundation models remain under-explored. We propose a framework that uses synthetic images to probe two applications of diffusion models, image editing and classification, for social bias. Using our framework, we uncover meaningful and significant inter-sectional social biases in Stable Diffusion, a state-of-the-art open-source text-to-image model. Our findings caution against the uninformed adoption of text-to-image foundation models for downstream tasks and services.
APA
Saravanan, A.P., Kocielnik, R., Jiang, R., Han, P. & Anandkumar, A.. (2023). Exploring Social Bias in Downstream Applications of Text-to-Image Foundation Models. Proceedings on "I Can't Believe It's Not Better: Failure Modes in the Age of Foundation Models" at NeurIPS 2023 Workshops, in Proceedings of Machine Learning Research 239:84-102 Available from https://proceedings.mlr.press/v239/saravanan23a.html.

Related Material