What type of study is this?

October 17, 2025Open Access

Laissez-Faire Harms: Algorithmic Biases in Generative Language Models (Extended Abstract)

View Full Paper

Ask AI

Bookmark

Generated texts from language models underrepresent minoritized individuals while reinforcing stereotypes.
Minoritized groups are overrepresented as subordinated characters, contrasting with dominant group narratives.
The analysis is based on responses from five leading generative language models examining societal harms.
Bias in character portrayal exceeds real-world demographics by drastic margins, indicating systemic issues.

No takes yet. Share an insight, caveat, or question.

Analysis reveals representational harms in synthetic texts from generative language models, suggesting bias impacts minoritized individuals.

Shieh et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

Analysis reveals representational harms in synthetic texts from generative language models, suggesting bias impacts minoritized individuals.