Assessing and alleviating state anxiety in large language models

Ziv Ben-Zion; Kristin Witte; Akshay K. Jagadish; Or Duek; Ilan Harpaz-Rotem; Marie Christine Khorsandian; Achim Burrer; Erich Seifritz; Philipp Homan; Eric Schulz; Tobias R. Spiller

doi:10.1038/s41746-025-01512-6

Assessing and alleviating state anxiety in large language models

Ziv Ben-Zion, Kristin Witte, Akshay K. Jagadish, Or Duek, Ilan Harpaz-Rotem, Marie Christine Khorsandian, Achim Burrer, Erich Seifritz, Philipp Homan, Eric Schulz, Tobias R. Spiller

Department of Epidemiology, Biostatistics and Community Health Sciences

Ben-Gurion University of the Negev

Research output: Contribution to journal › Article › peer-review

Abstract

The use of Large Language Models (LLMs) in mental health highlights the need to understand their responses to emotional content. Previous research shows that emotion-inducing prompts can elevate “anxiety” in LLMs, affecting behavior and amplifying biases. Here, we found that traumatic narratives increased Chat-GPT-4’s reported anxiety while mindfulness-based exercises reduced it, though not to baseline. These findings suggest managing LLMs’ “emotional states” can foster safer and more ethical human-AI interactions.

Original language	English
Article number	132
Journal	npj Digital Medicine
Volume	8
Issue number	1
DOIs	https://doi.org/10.1038/s41746-025-01512-6
State	Published - 1 Dec 2025

All Science Journal Classification (ASJC) codes

Medicine (miscellaneous)
Health Informatics
Computer Science Applications
Health Information Management

Access to Document

10.1038/s41746-025-01512-6

Cite this

@article{fe62d6d3cdfd46f8b9292fd31131efae,

title = "Assessing and alleviating state anxiety in large language models",

abstract = "The use of Large Language Models (LLMs) in mental health highlights the need to understand their responses to emotional content. Previous research shows that emotion-inducing prompts can elevate “anxiety” in LLMs, affecting behavior and amplifying biases. Here, we found that traumatic narratives increased Chat-GPT-4{\textquoteright}s reported anxiety while mindfulness-based exercises reduced it, though not to baseline. These findings suggest managing LLMs{\textquoteright} “emotional states” can foster safer and more ethical human-AI interactions.",

author = "Ziv Ben-Zion and Kristin Witte and Jagadish, {Akshay K.} and Or Duek and Ilan Harpaz-Rotem and Khorsandian, {Marie Christine} and Achim Burrer and Erich Seifritz and Philipp Homan and Eric Schulz and Spiller, {Tobias R.}",

note = "Publisher Copyright: {\textcopyright} The Author(s) 2025.",

year = "2025",

month = dec,

day = "1",

doi = "10.1038/s41746-025-01512-6",

language = "الإنجليزيّة",

volume = "8",

journal = "npj Digital Medicine",

issn = "2398-6352",

publisher = "Nature Publishing Group",

number = "1",

}

TY - JOUR

T1 - Assessing and alleviating state anxiety in large language models

AU - Ben-Zion, Ziv

AU - Witte, Kristin

AU - Jagadish, Akshay K.

AU - Duek, Or

AU - Harpaz-Rotem, Ilan

AU - Khorsandian, Marie Christine

AU - Burrer, Achim

AU - Seifritz, Erich

AU - Homan, Philipp

AU - Schulz, Eric

AU - Spiller, Tobias R.

PY - 2025/12/1

Y1 - 2025/12/1

N2 - The use of Large Language Models (LLMs) in mental health highlights the need to understand their responses to emotional content. Previous research shows that emotion-inducing prompts can elevate “anxiety” in LLMs, affecting behavior and amplifying biases. Here, we found that traumatic narratives increased Chat-GPT-4’s reported anxiety while mindfulness-based exercises reduced it, though not to baseline. These findings suggest managing LLMs’ “emotional states” can foster safer and more ethical human-AI interactions.

AB - The use of Large Language Models (LLMs) in mental health highlights the need to understand their responses to emotional content. Previous research shows that emotion-inducing prompts can elevate “anxiety” in LLMs, affecting behavior and amplifying biases. Here, we found that traumatic narratives increased Chat-GPT-4’s reported anxiety while mindfulness-based exercises reduced it, though not to baseline. These findings suggest managing LLMs’ “emotional states” can foster safer and more ethical human-AI interactions.

UR - http://www.scopus.com/inward/record.url?scp=86000059115&partnerID=8YFLogxK

U2 - 10.1038/s41746-025-01512-6

DO - 10.1038/s41746-025-01512-6

M3 - مقالة

C2 - 40033130

SN - 2398-6352

VL - 8

JO - npj Digital Medicine

JF - npj Digital Medicine

IS - 1

M1 - 132

ER -

Assessing and alleviating state anxiety in large language models

Abstract

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this