@inproceedings{dang-etal-2026-scenario,
title = "Scenario-based Probing and Steering Cultural Values in Large Language Models",
author = "Dang, Trung Duc Anh and
Kieu, Tung and
Masud, Sarah",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.858/",
pages = "18153--18174",
ISBN = "979-8-89176-462-0",
abstract = "Evaluation of cultural value alignment in Large Language Models (LLMs) typically relies on direct prompting with survey-style questions, which often elicit neutral or safety-aligned responses and fail to capture the model{'}s underlying cultural orientation. To overcome these challenges, we propose a framework for probing and steering latent cultural representations in LLMs along the two Inglehart{--}Welzel axes of the World Values Survey (WVS). By translating social value questions into scenario-based dilemmas, we extract token-level probabilities to measure implicit values and apply activation steering, optionally combined with country-conditioned prompting, to shift model behavior without retraining. Across three 4B-open-weight LLMs, four countries and three social domains, we find substantial variation in steerability. Importantly, we identify latent entanglement, where interventions along one Inglehart{--}Welzel axis induce a shift along the other. This entanglement is consistent with correlations present in human WVS data and persists across our experimental settings. Our work identifies empirical constraints on cultural alignment in LLMs."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="dang-etal-2026-scenario">
<titleInfo>
<title>Scenario-based Probing and Steering Cultural Values in Large Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Trung</namePart>
<namePart type="given">Duc</namePart>
<namePart type="given">Anh</namePart>
<namePart type="family">Dang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tung</namePart>
<namePart type="family">Kieu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sarah</namePart>
<namePart type="family">Masud</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Evaluation of cultural value alignment in Large Language Models (LLMs) typically relies on direct prompting with survey-style questions, which often elicit neutral or safety-aligned responses and fail to capture the model’s underlying cultural orientation. To overcome these challenges, we propose a framework for probing and steering latent cultural representations in LLMs along the two Inglehart–Welzel axes of the World Values Survey (WVS). By translating social value questions into scenario-based dilemmas, we extract token-level probabilities to measure implicit values and apply activation steering, optionally combined with country-conditioned prompting, to shift model behavior without retraining. Across three 4B-open-weight LLMs, four countries and three social domains, we find substantial variation in steerability. Importantly, we identify latent entanglement, where interventions along one Inglehart–Welzel axis induce a shift along the other. This entanglement is consistent with correlations present in human WVS data and persists across our experimental settings. Our work identifies empirical constraints on cultural alignment in LLMs.</abstract>
<identifier type="citekey">dang-etal-2026-scenario</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.858/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>18153</start>
<end>18174</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Scenario-based Probing and Steering Cultural Values in Large Language Models
%A Dang, Trung Duc Anh
%A Kieu, Tung
%A Masud, Sarah
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F dang-etal-2026-scenario
%X Evaluation of cultural value alignment in Large Language Models (LLMs) typically relies on direct prompting with survey-style questions, which often elicit neutral or safety-aligned responses and fail to capture the model’s underlying cultural orientation. To overcome these challenges, we propose a framework for probing and steering latent cultural representations in LLMs along the two Inglehart–Welzel axes of the World Values Survey (WVS). By translating social value questions into scenario-based dilemmas, we extract token-level probabilities to measure implicit values and apply activation steering, optionally combined with country-conditioned prompting, to shift model behavior without retraining. Across three 4B-open-weight LLMs, four countries and three social domains, we find substantial variation in steerability. Importantly, we identify latent entanglement, where interventions along one Inglehart–Welzel axis induce a shift along the other. This entanglement is consistent with correlations present in human WVS data and persists across our experimental settings. Our work identifies empirical constraints on cultural alignment in LLMs.
%U https://aclanthology.org/2026.findings-emnlp.858/
%P 18153-18174
Markdown (Informal)
[Scenario-based Probing and Steering Cultural Values in Large Language Models](https://aclanthology.org/2026.findings-emnlp.858/) (Dang et al., Findings 2026)
ACL