Harmful Stereotyping of Non-Cisgendered People via Text-to-Image Systems

July 3, 2023

Text-to-image systems like DALL-E have raised concerns due to their alleged bias and offensive output when generating images based on non-cisgender identities. These systems tend to produce stereotypical, sexualized, and discriminatory representations in response to gender identity terms such as 'trans', 'nonbinary', or 'queer', indicating systemic biases that require attention.

For those interested in shaping the future of trustworthy AI and addressing these systemic issues, we invite you to join us in the HISPI Project Cerebellum TAIM initiative. Together, we can Measure, Map, Manage, and Govern AI to promote safe and secure practices.
JOIN US

Matched TAIM controls

Suggested mapping from embedding similarity (not a formal assessment). Browse all TAIM controls

MAP 1.6 — similarity 0.674, rank 1. TAIM detail and related incidents →
MAP 4.1 — similarity 0.665, rank 2. TAIM detail and related incidents →
MAP 1.1 — similarity 0.657, rank 3. TAIM detail and related incidents →

Alleged deployer: dall-e
Alleged developer: openai
Alleged harmed parties: non-cisgender-individuals, lgbtq+-community

Source

Data from the AI Incident Database (AIID). Cite this incident: https://incidentdatabase.ai/cite/579

Data source

Incident data is from the AI Incident Database (AIID).

When citing the database as a whole, please use:

McGregor, S. (2021) Preventing Repeated Real World AI Failures by Cataloging Incidents: The AI Incident Database. In Proceedings of the Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (IAAI-21). Virtual Conference.

Pre-print on arXiv · Database snapshots & citation guide

We use weekly snapshots of the AIID for stable reference. For the official suggested citation of a specific incident, use the “Cite this incident” link on each incident page.