@inproceedings{tan-lee-2025-unmasking,
title = "Unmasking Implicit Bias: Evaluating Persona-Prompted {LLM} Responses in Power-Disparate Social Scenarios",
author = "Tan, Bryan Chen Zhengyu and
Lee, Roy Ka-Wei",
editor = "Chiruzzo, Luis and
Ritter, Alan and
Wang, Lu",
booktitle = "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)",
month = apr,
year = "2025",
address = "Albuquerque, New Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.naacl-long.50/",
doi = "10.18653/v1/2025.naacl-long.50",
pages = "1075--1108",
ISBN = "979-8-89176-189-6",
abstract = "Large language models (LLMs) have demonstrated remarkable capabilities in simulating human behaviour and social intelligence. However, they risk perpetuating societal biases, especially when demographic information is involved. We introduce a novel framework using cosine distance to measure semantic shifts in responses and an LLM-judged Preference Win Rate (WR) to assess how demographic prompts affect response quality across power-disparate social scenarios. Evaluating five LLMs over 100 diverse social scenarios and nine demographic axes, our findings suggest a ``default persona'' bias toward middle-aged, able-bodied, native-born, Caucasian, atheistic males with centrist views. Moreover, interactions involving specific demographics are associated with lower-quality responses. Lastly, the presence of power disparities increases variability in response semantics and quality across demographic groups, suggesting that implicit biases may be heightened under power-imbalanced conditions. These insights expose the demographic biases inherent in LLMs and offer potential paths toward future bias mitigation efforts in LLMs."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="tan-lee-2025-unmasking">
<titleInfo>
<title>Unmasking Implicit Bias: Evaluating Persona-Prompted LLM Responses in Power-Disparate Social Scenarios</title>
</titleInfo>
<name type="personal">
<namePart type="given">Bryan</namePart>
<namePart type="given">Chen</namePart>
<namePart type="given">Zhengyu</namePart>
<namePart type="family">Tan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Roy</namePart>
<namePart type="given">Ka-Wei</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-04</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Luis</namePart>
<namePart type="family">Chiruzzo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alan</namePart>
<namePart type="family">Ritter</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Albuquerque, New Mexico</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-189-6</identifier>
</relatedItem>
<abstract>Large language models (LLMs) have demonstrated remarkable capabilities in simulating human behaviour and social intelligence. However, they risk perpetuating societal biases, especially when demographic information is involved. We introduce a novel framework using cosine distance to measure semantic shifts in responses and an LLM-judged Preference Win Rate (WR) to assess how demographic prompts affect response quality across power-disparate social scenarios. Evaluating five LLMs over 100 diverse social scenarios and nine demographic axes, our findings suggest a “default persona” bias toward middle-aged, able-bodied, native-born, Caucasian, atheistic males with centrist views. Moreover, interactions involving specific demographics are associated with lower-quality responses. Lastly, the presence of power disparities increases variability in response semantics and quality across demographic groups, suggesting that implicit biases may be heightened under power-imbalanced conditions. These insights expose the demographic biases inherent in LLMs and offer potential paths toward future bias mitigation efforts in LLMs.</abstract>
<identifier type="citekey">tan-lee-2025-unmasking</identifier>
<identifier type="doi">10.18653/v1/2025.naacl-long.50</identifier>
<location>
<url>https://aclanthology.org/2025.naacl-long.50/</url>
</location>
<part>
<date>2025-04</date>
<extent unit="page">
<start>1075</start>
<end>1108</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Unmasking Implicit Bias: Evaluating Persona-Prompted LLM Responses in Power-Disparate Social Scenarios
%A Tan, Bryan Chen Zhengyu
%A Lee, Roy Ka-Wei
%Y Chiruzzo, Luis
%Y Ritter, Alan
%Y Wang, Lu
%S Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)
%D 2025
%8 April
%I Association for Computational Linguistics
%C Albuquerque, New Mexico
%@ 979-8-89176-189-6
%F tan-lee-2025-unmasking
%X Large language models (LLMs) have demonstrated remarkable capabilities in simulating human behaviour and social intelligence. However, they risk perpetuating societal biases, especially when demographic information is involved. We introduce a novel framework using cosine distance to measure semantic shifts in responses and an LLM-judged Preference Win Rate (WR) to assess how demographic prompts affect response quality across power-disparate social scenarios. Evaluating five LLMs over 100 diverse social scenarios and nine demographic axes, our findings suggest a “default persona” bias toward middle-aged, able-bodied, native-born, Caucasian, atheistic males with centrist views. Moreover, interactions involving specific demographics are associated with lower-quality responses. Lastly, the presence of power disparities increases variability in response semantics and quality across demographic groups, suggesting that implicit biases may be heightened under power-imbalanced conditions. These insights expose the demographic biases inherent in LLMs and offer potential paths toward future bias mitigation efforts in LLMs.
%R 10.18653/v1/2025.naacl-long.50
%U https://aclanthology.org/2025.naacl-long.50/
%U https://doi.org/10.18653/v1/2025.naacl-long.50
%P 1075-1108
Markdown (Informal)
[Unmasking Implicit Bias: Evaluating Persona-Prompted LLM Responses in Power-Disparate Social Scenarios](https://aclanthology.org/2025.naacl-long.50/) (Tan & Lee, NAACL 2025)
ACL