@inproceedings{dutta-etal-2026-auditing,
title = "Auditing {LLM} Responses to Harmful Stereotypes Targeting Mental Health Groups",
author = "Dutta, Arka and
Magu, Rijul and
Kim, Sean and
Yoon, Seohee and
De Choudhury, Munmun and
KhudaBukhsh, Ashiqur R.",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-acl.2010/",
pages = "40435--40452",
ISBN = "979-8-89176-395-1",
abstract = "Large Language Models (LLMs) can exhibit imbalanced biases against vulnerable groups, but how they rationalize stereotypes and rights restrictions targeting mental health entities remains underexplored. We audit a broad suite of open-weight LLMs on stereotype-justification prompts tied to mental health identities. We find that several widely used models endorse harmful stereotypes when explicitly asked to justify them, with endorsement varying across model families, versions, and mental health conditions. Finally, we show that widely used harmful-content evaluation and moderation frameworks often miss these nuanced, discriminatory responses, highlighting a gap in current AI safety evaluation for mental health groups."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="dutta-etal-2026-auditing">
<titleInfo>
<title>Auditing LLM Responses to Harmful Stereotypes Targeting Mental Health Groups</title>
</titleInfo>
<name type="personal">
<namePart type="given">Arka</namePart>
<namePart type="family">Dutta</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rijul</namePart>
<namePart type="family">Magu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sean</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Seohee</namePart>
<namePart type="family">Yoon</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Munmun</namePart>
<namePart type="family">De Choudhury</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ashiqur</namePart>
<namePart type="given">R</namePart>
<namePart type="family">KhudaBukhsh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Liakata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Viviane</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Jurgens</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-395-1</identifier>
</relatedItem>
<abstract>Large Language Models (LLMs) can exhibit imbalanced biases against vulnerable groups, but how they rationalize stereotypes and rights restrictions targeting mental health entities remains underexplored. We audit a broad suite of open-weight LLMs on stereotype-justification prompts tied to mental health identities. We find that several widely used models endorse harmful stereotypes when explicitly asked to justify them, with endorsement varying across model families, versions, and mental health conditions. Finally, we show that widely used harmful-content evaluation and moderation frameworks often miss these nuanced, discriminatory responses, highlighting a gap in current AI safety evaluation for mental health groups.</abstract>
<identifier type="citekey">dutta-etal-2026-auditing</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.2010/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>40435</start>
<end>40452</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Auditing LLM Responses to Harmful Stereotypes Targeting Mental Health Groups
%A Dutta, Arka
%A Magu, Rijul
%A Kim, Sean
%A Yoon, Seohee
%A De Choudhury, Munmun
%A KhudaBukhsh, Ashiqur R.
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Findings of the Association for Computational Linguistics: ACL 2026
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-395-1
%F dutta-etal-2026-auditing
%X Large Language Models (LLMs) can exhibit imbalanced biases against vulnerable groups, but how they rationalize stereotypes and rights restrictions targeting mental health entities remains underexplored. We audit a broad suite of open-weight LLMs on stereotype-justification prompts tied to mental health identities. We find that several widely used models endorse harmful stereotypes when explicitly asked to justify them, with endorsement varying across model families, versions, and mental health conditions. Finally, we show that widely used harmful-content evaluation and moderation frameworks often miss these nuanced, discriminatory responses, highlighting a gap in current AI safety evaluation for mental health groups.
%U https://aclanthology.org/2026.findings-acl.2010/
%P 40435-40452
Markdown (Informal)
[Auditing LLM Responses to Harmful Stereotypes Targeting Mental Health Groups](https://aclanthology.org/2026.findings-acl.2010/) (Dutta et al., Findings 2026)
ACL