@inproceedings{le-beux-etal-2026-afristereo,
title = "{A}fri{S}tereo: A Culturally Grounded Dataset for Evaluating Stereotypical Bias in Large Language Models",
author = "Le Beux, Yann and
Audu, Oluchi and
Ankeli, Oche David and
Balakrishnan, Dhananjay and
Weya, Melissah and
Ralaiarinosy, Marie Daniella and
Ezeani, Ignatius",
editor = "Piperidis, Stelios and
Bel, N{\'u}ria and
van den Heuvel, Henk and
Ide, Nancy and
Krek, Simon and
Toral, Antonio",
booktitle = "Proceedings of the Fifteenth Language Resources and Evaluation Conference",
month = may,
year = "2026",
address = "Palma de Mallorca, Spain",
publisher = "ELRA Language Resource Association",
url = "https://aclanthology.org/2026.lrec-1.18/",
doi = "10.63317/58oaqcxpogdy",
pages = "266--280",
abstract = "Existing AI bias evaluation benchmarks largely reflect Western perspectives, leaving African contexts underrepresented and enabling harmful stereotypes in applications across various domains. To address this gap, we introduce AfriStereo, the first open-source African stereotype dataset and evaluation framework grounded in local socio-cultural contexts. Through community engaged efforts across Senegal, Kenya, and Nigeria, we collect 1,163 stereotypes spanning gender, ethnicity, religion, age, and profession. Using few-shot prompting with human-in-the-loop validation, we augment the dataset to over 5,000 stereotype{--}antistereotype pairs. Entries are validated through semantic clustering and manual annotation by culturally informed reviewers. Preliminary evaluation of language models reveals that nine of eleven models exhibit statistically significant bias in our setup, with Bias Preference Ratios (BPR) ranging from 0.63 to 0.78 (p {\ensuremath{\leq}} 0.05), indicating systematic preferences for stereotypes over antistereotypes, particularly across age, profession, and gender dimensions. Domain-specific models appear to show weaker bias in our setup, suggesting task-specific training may mitigate some associations. Looking ahead, AfriStereo opens pathways for future research on culturally grounded bias evaluation and mitigation, offering key methodologies for the AI community on building more equitable, context-aware, and globally inclusive NLP technologies."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="le-beux-etal-2026-afristereo">
<titleInfo>
<title>AfriStereo: A Culturally Grounded Dataset for Evaluating Stereotypical Bias in Large Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yann</namePart>
<namePart type="family">Le Beux</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Oluchi</namePart>
<namePart type="family">Audu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Oche</namePart>
<namePart type="given">David</namePart>
<namePart type="family">Ankeli</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dhananjay</namePart>
<namePart type="family">Balakrishnan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Melissah</namePart>
<namePart type="family">Weya</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marie</namePart>
<namePart type="given">Daniella</namePart>
<namePart type="family">Ralaiarinosy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ignatius</namePart>
<namePart type="family">Ezeani</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 Fifteenth Language Resources and Evaluation Conference</title>
</titleInfo>
<name type="personal">
<namePart type="given">Stelios</namePart>
<namePart type="family">Piperidis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Núria</namePart>
<namePart type="family">Bel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Henk</namePart>
<namePart type="family">van den Heuvel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nancy</namePart>
<namePart type="family">Ide</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Simon</namePart>
<namePart type="family">Krek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Antonio</namePart>
<namePart type="family">Toral</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>ELRA Language Resource Association</publisher>
<place>
<placeTerm type="text">Palma de Mallorca, Spain</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Existing AI bias evaluation benchmarks largely reflect Western perspectives, leaving African contexts underrepresented and enabling harmful stereotypes in applications across various domains. To address this gap, we introduce AfriStereo, the first open-source African stereotype dataset and evaluation framework grounded in local socio-cultural contexts. Through community engaged efforts across Senegal, Kenya, and Nigeria, we collect 1,163 stereotypes spanning gender, ethnicity, religion, age, and profession. Using few-shot prompting with human-in-the-loop validation, we augment the dataset to over 5,000 stereotype–antistereotype pairs. Entries are validated through semantic clustering and manual annotation by culturally informed reviewers. Preliminary evaluation of language models reveals that nine of eleven models exhibit statistically significant bias in our setup, with Bias Preference Ratios (BPR) ranging from 0.63 to 0.78 (p \ensuremathłeq 0.05), indicating systematic preferences for stereotypes over antistereotypes, particularly across age, profession, and gender dimensions. Domain-specific models appear to show weaker bias in our setup, suggesting task-specific training may mitigate some associations. Looking ahead, AfriStereo opens pathways for future research on culturally grounded bias evaluation and mitigation, offering key methodologies for the AI community on building more equitable, context-aware, and globally inclusive NLP technologies.</abstract>
<identifier type="citekey">le-beux-etal-2026-afristereo</identifier>
<identifier type="doi">10.63317/58oaqcxpogdy</identifier>
<location>
<url>https://aclanthology.org/2026.lrec-1.18/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>266</start>
<end>280</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T AfriStereo: A Culturally Grounded Dataset for Evaluating Stereotypical Bias in Large Language Models
%A Le Beux, Yann
%A Audu, Oluchi
%A Ankeli, Oche David
%A Balakrishnan, Dhananjay
%A Weya, Melissah
%A Ralaiarinosy, Marie Daniella
%A Ezeani, Ignatius
%Y Piperidis, Stelios
%Y Bel, Núria
%Y van den Heuvel, Henk
%Y Ide, Nancy
%Y Krek, Simon
%Y Toral, Antonio
%S Proceedings of the Fifteenth Language Resources and Evaluation Conference
%D 2026
%8 May
%I ELRA Language Resource Association
%C Palma de Mallorca, Spain
%F le-beux-etal-2026-afristereo
%X Existing AI bias evaluation benchmarks largely reflect Western perspectives, leaving African contexts underrepresented and enabling harmful stereotypes in applications across various domains. To address this gap, we introduce AfriStereo, the first open-source African stereotype dataset and evaluation framework grounded in local socio-cultural contexts. Through community engaged efforts across Senegal, Kenya, and Nigeria, we collect 1,163 stereotypes spanning gender, ethnicity, religion, age, and profession. Using few-shot prompting with human-in-the-loop validation, we augment the dataset to over 5,000 stereotype–antistereotype pairs. Entries are validated through semantic clustering and manual annotation by culturally informed reviewers. Preliminary evaluation of language models reveals that nine of eleven models exhibit statistically significant bias in our setup, with Bias Preference Ratios (BPR) ranging from 0.63 to 0.78 (p \ensuremathłeq 0.05), indicating systematic preferences for stereotypes over antistereotypes, particularly across age, profession, and gender dimensions. Domain-specific models appear to show weaker bias in our setup, suggesting task-specific training may mitigate some associations. Looking ahead, AfriStereo opens pathways for future research on culturally grounded bias evaluation and mitigation, offering key methodologies for the AI community on building more equitable, context-aware, and globally inclusive NLP technologies.
%R 10.63317/58oaqcxpogdy
%U https://aclanthology.org/2026.lrec-1.18/
%U https://doi.org/10.63317/58oaqcxpogdy
%P 266-280
Markdown (Informal)
[AfriStereo: A Culturally Grounded Dataset for Evaluating Stereotypical Bias in Large Language Models](https://aclanthology.org/2026.lrec-1.18/) (Le Beux et al., LREC 2026)
ACL