EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Auditing and instructing text-to-image generation models on fairness

Felix Friedrich iD, Manuel Brack, Lukas Struppek, Dominik Hintersdorf, Patrick Schramowski, Sasha Luccioni, Kristian Kersting

AI and Ethics · 2025-06 · 5 (3) · 2103-2123

DOI: 10.1007/s43681-024-00531-5 ↗

Kurz gefasst

Fachartikel in der Zeitschrift AI and Ethics, online erschienen am 01.08.2024. Verfasst von Felix Friedrich und 6 weiteren Autor:innen. Beteiligte Einrichtung: Artificial Intelligence and Machine Learning Lab, Technical University of Darmstadt, Darmstadt, Germany. Die Arbeit wurde bisher 55-mal zitiert (Crossref) und nennt 63 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: HORIZON EUROPE European Research Council, Deutsches Forschungszentrum für Künstliche Intelligenz, Hessisches Ministerium für Wissenschaft und Kunst.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Generative AI models have recently achieved astonishing results in quality and are consequently employed in a fast-growing number of applications. However, since they are highly data-driven, relying on billion-sized datasets randomly scraped from the internet, they also suffer from degenerated and biased human behavior, as we demonstrate. In fact, they may even reinforce such biases. To not only uncover but also combat these undesired effects, we present a novel strategy, called Fair Diffusion , to attenuate biases during the deployment of generative text-to-image models. Specifically, we demonstrate shifting a bias in any direction based on human instructions yielding arbitrary proportions for, e.g., identity groups. As our empirical evaluation demonstrates, this introduced control enables instructing generative image models on fairness, requiring no data filtering nor additional training.

Schlagwörter

Stichwörter der Autor:innen

FairnessModel auditModel debiasingStable diffusionText-guided image generationText-to-image synthesis

Autor:innen und Institutionen

  • Artificial Intelligence and Machine Learning Lab, Technical University of Darmstadt, Darmstadt, Germany.
  • Hessian Center for Artificial Intelligence (hessian.AI), Darmstadt, Germany.
  • German Research Center for Artificial Intelligence (DFKI), Darmstadt, Germany.
  • LAION, Online, Hamburg, Germany.
  • Huggingface, Montreal, Canada.
  • Centre for Cognitive Science, Technical University of Darmstadt, Darmstadt, Germany.

ORCID-Profile: Felix Friedrich

Förderung

  • HORIZON EUROPE European Research Council
  • Deutsches Forschungszentrum für Künstliche Intelligenz
  • Hessisches Ministerium für Wissenschaft und Kunst
  • Technische Universität Darmstadt

Zitieren

APA

Friedrich, F., Brack, M., Struppek, L., Hintersdorf, D., Schramowski, P., Luccioni, S., & Kersting, K. (2025). Auditing and instructing text-to-image generation models on fairness. AI and Ethics, 5(3), 2103-2123. https://doi.org/10.1007/s43681-024-00531-5

Vancouver

Friedrich F, Brack M, Struppek L, Hintersdorf D, Schramowski P, Luccioni S, et al. Auditing and instructing text-to-image generation models on fairness. AI and Ethics. 2025;5(3):2103-2123. doi:10.1007/s43681-024-00531-5

BibTeX
@article{Friedrich2025,
  author = {Friedrich, Felix and Brack, Manuel and Struppek, Lukas and Hintersdorf, Dominik and Schramowski, Patrick and Luccioni, Sasha and Kersting, Kristian},
  title = {Auditing and instructing text-to-image generation models on fairness},
  journal = {AI and Ethics},
  year = {2025},
  volume = {5},
  number = {3},
  pages = {2103--2123},
  doi = {10.1007/s43681-024-00531-5}
}
Kostenlos starten →