@inproceedings{majumdar-etal-2026-evaluating,
title = "Evaluating {LLM}s for Detecting Demographic-Targeted Social Bias: A Comprehensive Benchmark Study",
author = "Majumdar, Ayan and
Chen, Feihao and
Li, Jinghui and
Wang, Xiaozhen",
editor = "Pranav, A and
Basile, Valerio and
Falk, Neele and
Jurgens, David and
Lapesa, Gabriella and
Lauscher, Anne and
Lo, Soda Marem",
booktitle = "Proceedings of the Second Workshop of Identity Aware {AI}",
month = may,
year = "2026",
address = "Palma de Mallorca, Spain",
publisher = "European Language Resources Association",
url = "https://aclanthology.org/2026.iaai-1.5/",
doi = "10.63317/3jejtg2hfj3t",
pages = "47--65",
abstract = "Large-scale web-scraped text corpora used to train general-purpose AI models often contain harmful demographic-targeted social biases, creating a regulatory need for data auditing and developing scalable bias-detection methods. Although prior work has investigated biases in text datasets and related detection methods, these studies remain narrow in scope. They typically focus on a single content type (e.g., hate speech), cover limited demographic axes, overlook biases affecting multiple demographics simultaneously, and analyze limited techniques. Consequently, practitioners lack a holistic understanding of the strengths and limitations of recent large language models (LLMs) for automated bias detection. In this study, we conduct a comprehensive benchmark study on English texts to assess the ability of LLMs in detecting demographic-targeted social biases. To align with regulatory requirements, we frame bias detection as a multi-label task of detecting targeted identities using a demographic-focused taxonomy. We then systematically evaluate models across scales and techniques, including prompting, in-context learning, and fine-tuning. Using twelve datasets spanning diverse content types and demographics, our study demonstrates the promise of fine-tuned smaller models for scalable detection. However, our analyses also expose persistent gaps across identity axes and multi-demographic targeted biases, underscoring the need for more effective and scalable detection frameworks."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="majumdar-etal-2026-evaluating">
<titleInfo>
<title>Evaluating LLMs for Detecting Demographic-Targeted Social Bias: A Comprehensive Benchmark Study</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ayan</namePart>
<namePart type="family">Majumdar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Feihao</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jinghui</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiaozhen</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Second Workshop of Identity Aware AI</title>
</titleInfo>
<name type="personal">
<namePart type="given">A</namePart>
<namePart type="family">Pranav</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Valerio</namePart>
<namePart type="family">Basile</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Neele</namePart>
<namePart type="family">Falk</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>
<name type="personal">
<namePart type="given">Gabriella</namePart>
<namePart type="family">Lapesa</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anne</namePart>
<namePart type="family">Lauscher</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Soda</namePart>
<namePart type="given">Marem</namePart>
<namePart type="family">Lo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>European Language Resources Association</publisher>
<place>
<placeTerm type="text">Palma de Mallorca, Spain</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Large-scale web-scraped text corpora used to train general-purpose AI models often contain harmful demographic-targeted social biases, creating a regulatory need for data auditing and developing scalable bias-detection methods. Although prior work has investigated biases in text datasets and related detection methods, these studies remain narrow in scope. They typically focus on a single content type (e.g., hate speech), cover limited demographic axes, overlook biases affecting multiple demographics simultaneously, and analyze limited techniques. Consequently, practitioners lack a holistic understanding of the strengths and limitations of recent large language models (LLMs) for automated bias detection. In this study, we conduct a comprehensive benchmark study on English texts to assess the ability of LLMs in detecting demographic-targeted social biases. To align with regulatory requirements, we frame bias detection as a multi-label task of detecting targeted identities using a demographic-focused taxonomy. We then systematically evaluate models across scales and techniques, including prompting, in-context learning, and fine-tuning. Using twelve datasets spanning diverse content types and demographics, our study demonstrates the promise of fine-tuned smaller models for scalable detection. However, our analyses also expose persistent gaps across identity axes and multi-demographic targeted biases, underscoring the need for more effective and scalable detection frameworks.</abstract>
<identifier type="citekey">majumdar-etal-2026-evaluating</identifier>
<identifier type="doi">10.63317/3jejtg2hfj3t</identifier>
<location>
<url>https://aclanthology.org/2026.iaai-1.5/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>47</start>
<end>65</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Evaluating LLMs for Detecting Demographic-Targeted Social Bias: A Comprehensive Benchmark Study
%A Majumdar, Ayan
%A Chen, Feihao
%A Li, Jinghui
%A Wang, Xiaozhen
%Y Pranav, A.
%Y Basile, Valerio
%Y Falk, Neele
%Y Jurgens, David
%Y Lapesa, Gabriella
%Y Lauscher, Anne
%Y Lo, Soda Marem
%S Proceedings of the Second Workshop of Identity Aware AI
%D 2026
%8 May
%I European Language Resources Association
%C Palma de Mallorca, Spain
%F majumdar-etal-2026-evaluating
%X Large-scale web-scraped text corpora used to train general-purpose AI models often contain harmful demographic-targeted social biases, creating a regulatory need for data auditing and developing scalable bias-detection methods. Although prior work has investigated biases in text datasets and related detection methods, these studies remain narrow in scope. They typically focus on a single content type (e.g., hate speech), cover limited demographic axes, overlook biases affecting multiple demographics simultaneously, and analyze limited techniques. Consequently, practitioners lack a holistic understanding of the strengths and limitations of recent large language models (LLMs) for automated bias detection. In this study, we conduct a comprehensive benchmark study on English texts to assess the ability of LLMs in detecting demographic-targeted social biases. To align with regulatory requirements, we frame bias detection as a multi-label task of detecting targeted identities using a demographic-focused taxonomy. We then systematically evaluate models across scales and techniques, including prompting, in-context learning, and fine-tuning. Using twelve datasets spanning diverse content types and demographics, our study demonstrates the promise of fine-tuned smaller models for scalable detection. However, our analyses also expose persistent gaps across identity axes and multi-demographic targeted biases, underscoring the need for more effective and scalable detection frameworks.
%R 10.63317/3jejtg2hfj3t
%U https://aclanthology.org/2026.iaai-1.5/
%U https://doi.org/10.63317/3jejtg2hfj3t
%P 47-65
Markdown (Informal)
[Evaluating LLMs for Detecting Demographic-Targeted Social Bias: A Comprehensive Benchmark Study](https://aclanthology.org/2026.iaai-1.5/) (Majumdar et al., iaai 2026)
ACL