What type of study is this?

This is a Quantitative Study study.

October 3, 2025Open Access

Bias in Large Language Models for Mental Health: Evidence from Vignette-Based Evaluation Across Nine Models

View Full Paper

Ask AI

Bookmark

Models generated biased responses, demonstrating notable lexical and semantic differences from human expert responses.
Automated metrics like BERTScoreF and ROUGE-L were used to evaluate the degree of bias across nine large language models.
Interaction effects of symptom type and severity revealed weaker concordance between LLMs and human experts, indicating biases.
The study emphasizes the need for human expert oversight in developing and applying large language models for mental health.

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

Analysis uncovers bias in large language models' responses to mental health vignettes, suggesting oversight is crucial.

Wei et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark