AI representations of mental health stigma: Quantifying the gap between AI estimations and human perceptions across gender groups using GPT-based LLMs
Social Science & Medicine · 2026-12 · 410 · 119855
DOI: 10.1016/j.socscimed.2026.119855 ↗Kurz gefasst
Fachartikel in der Zeitschrift Social Science & Medicine, 2026-12. Verfasst von Soojong Kim und 3 weiteren Autor:innen. Beteiligte Einrichtung: Department of Communication, University of California Davis, United States. Electronic address: sjokim@ucdavis.edu. Hauptthemen laut MEDLINE: Social Stigma, Artificial Intelligence, Mental Disorders. Die Arbeit nennt 66 Literaturangaben.
Abstract (Originalsprache)
The increasing use of artificial intelligence (AI) in mental health applications raises concerns about its potential to internalize and perpetuate societal biases. This research evaluates both disparities and misalignment in how large language models (LLMs), GPT-4 and GPT-5, represent perceptions of mental health across gender groups. The research comprises two studies. Study 1 applied widely adopted measures of Social Distance, Stereotypes, and Attitudes Toward Seeking Help to assess whether the AI models depict systematic gender differences. Study 2 compared AI estimations of Personal, Self, and Perceived Stigma with large-scale real-world human data from the Healthy Minds Study (N = 76,406) to evaluate alignment between AI estimations and human perceptions. Results from Study 1 showed the AI models consistently portrayed men as having the greatest desire for social distance from someone with mental health problems, relatively higher Stereotypes scores, and the least favorable attitudes toward help-seeking, compared with women and nonbinary individuals. Study 2 confirmed that this ordinal pattern partially aligned with human-reported data for Personal and Self Stigma. However, the AI models systematically underestimated Personal Stigma and Perceived Stigma while substantially overestimating Self Stigma for all gender groups. These findings reveal that, although the AI models have learned relative gender-based patterns reflecting human experiences to a certain extent, they function as distorted portrayals, highlighting the need for rigorous, context-specific validation against real-world data to ensure AI's safe and equitable deployment.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Communication, University of California Davis, United States. Electronic address: sjokim@ucdavis.edu.
- Department of Communication, University of Massachusetts Boston, United States.
- Department of Media, College of Politics and Economics, Kyung Hee University, South Korea.
- Department of Communication, University of California Davis, United States.
Zitieren
APA
Kim, S., Kim, H. M., Kim, K., & Zhu, D. (2026). AI representations of mental health stigma: Quantifying the gap between AI estimations and human perceptions across gender groups using GPT-based LLMs. Social Science & Medicine, 410, 119855. https://doi.org/10.1016/j.socscimed.2026.119855
Vancouver
Kim S, Kim HM, Kim K, Zhu D. AI representations of mental health stigma: Quantifying the gap between AI estimations and human perceptions across gender groups using GPT-based LLMs. Social Science & Medicine. 2026;410:119855. doi:10.1016/j.socscimed.2026.119855
BibTeX
@article{Kim2026,
author = {Kim, Soojong and Kim, Hye Min and Kim, Kwanho and Zhu, Dongdong},
title = {AI representations of mental health stigma: Quantifying the gap between AI estimations and human perceptions across gender groups using GPT-based LLMs},
journal = {Social Science & Medicine},
year = {2026},
volume = {410},
pages = {119855},
doi = {10.1016/j.socscimed.2026.119855}
}
EUVIMED