Saved in:
Bibliographic Details
Main Authors: Myers, Quintin, Gao, Yanjun
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.20822
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909660217344000
author Myers, Quintin
Gao, Yanjun
author_facet Myers, Quintin
Gao, Yanjun
contents Large language models (LLMs) are increasingly proposed for detecting and responding to violent content online, yet their ability to reason about morally ambiguous, real-world scenarios remains underexamined. We present the first study to evaluate LLMs using a validated social science instrument designed to measure human response to everyday conflict, namely the Violent Behavior Vignette Questionnaire (VBVQ). To assess potential bias, we introduce persona-based prompting that varies race, age, and geographic identity within the United States. Six LLMs developed across different geopolitical and organizational contexts are evaluated under a unified zero-shot setting. Our study reveals two key findings: (1) LLMs surface-level text generation often diverges from their internal preference for violent responses; (2) their violent tendencies vary across demographics, frequently contradicting established findings in criminology, social science, and psychology.
format Preprint
id arxiv_https___arxiv_org_abs_2506_20822
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Uncovering Hidden Violent Tendencies in LLMs: A Demographic Analysis via Behavioral Vignettes
Myers, Quintin
Gao, Yanjun
Computation and Language
Artificial Intelligence
Large language models (LLMs) are increasingly proposed for detecting and responding to violent content online, yet their ability to reason about morally ambiguous, real-world scenarios remains underexamined. We present the first study to evaluate LLMs using a validated social science instrument designed to measure human response to everyday conflict, namely the Violent Behavior Vignette Questionnaire (VBVQ). To assess potential bias, we introduce persona-based prompting that varies race, age, and geographic identity within the United States. Six LLMs developed across different geopolitical and organizational contexts are evaluated under a unified zero-shot setting. Our study reveals two key findings: (1) LLMs surface-level text generation often diverges from their internal preference for violent responses; (2) their violent tendencies vary across demographics, frequently contradicting established findings in criminology, social science, and psychology.
title Uncovering Hidden Violent Tendencies in LLMs: A Demographic Analysis via Behavioral Vignettes
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2506.20822