LLMs and Cultural Values: The Impact of Prompt Language and Explicit Cultural Framing.
Saved in:
| Title: | LLMs and Cultural Values: The Impact of Prompt Language and Explicit Cultural Framing. |
|---|---|
| Authors: | Bulté, Bram1 (AUTHOR) bram.bulte@vub.be, Terryn, Ayla Rigouts2 (AUTHOR) ayla.rigouts.terryn@umontreal.ca |
| Source: | Computational Linguistics. Jun2026, Vol. 52 Issue 2, p407-494. 88p. |
| Subjects: | Cultural values, Algorithmic bias, Natural language processing, Social values, Language & languages, Language models, Social attitudes |
| Geographic Terms: | Netherlands, United States, Japan, Germany |
| Abstract: | Large language models (LLMs) are rapidly being adopted by users across the globe, who interact with them in a diverse range of languages. At the same time, there are well-documented imbalances in the training data and optimization objectives of this technology, raising doubts as to whether LLMs can accurately represent the cultural diversity of their broad user base. In this study, we look at LLMs and cultural values in particular, and examine how prompt language and cultural framing influence model responses and their alignment with human values in different countries. We do so by probing 10 LLMs with 63 items from the Hofstede Values Survey Module and World Values Survey, translated into 11 languages, and formulated as prompts with and without different explicit cultural perspectives. Our study confirms that both prompt language and cultural perspective produce variation in LLM outputs, but with an important caveat: While targeted prompting can, to a certain extent, steer LLM responses in the direction of the predominant values of the corresponding countries, it does not overcome the models' systematic bias toward the values associated with a restricted set of countries in our dataset: the Netherlands, Germany, the United States, and Japan. All tested models, regardless of their origin, exhibit remarkably similar patterns: They produce fairly neutral responses on most topics, with selective progressive stances on issues such as social tolerance. Alignment with cultural values of human respondents is improved more with an explicit cultural perspective than with a targeted prompt language. Unexpectedly, combining both approaches is no more effective than cultural framing with an English prompt. These findings reveal that LLMs occupy an uncomfortable middle ground: They are responsive enough to changes in prompts to produce variation, but they are also too firmly anchored to specific cultural defaults to adequately represent cultural diversity. [ABSTRACT FROM AUTHOR] |
| Copyright of Computational Linguistics is the property of MIT Press and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| FullText | Links: – Type: pdflink Text: Availability: 1 |
|---|---|
| Header | DbId: egs DbLabel: Engineering Source An: 194945038 AccessLevel: 6 PubType: Academic Journal PubTypeId: academicJournal PreciseRelevancyScore: 0 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: LLMs and Cultural Values: The Impact of Prompt Language and Explicit Cultural Framing. – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Bulté%2C+Bram%22">Bulté, Bram</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> bram.bulte@vub.be</i><br /><searchLink fieldCode="AR" term="%22Terryn%2C+Ayla+Rigouts%22">Terryn, Ayla Rigouts</searchLink><relatesTo>2</relatesTo> (AUTHOR)<i> ayla.rigouts.terryn@umontreal.ca</i> – Name: TitleSource Label: Source Group: Src Data: <searchLink fieldCode="JN" term="%22Computational+Linguistics%22">Computational Linguistics</searchLink>. Jun2026, Vol. 52 Issue 2, p407-494. 88p. – Name: Subject Label: Subjects Group: Su Data: <searchLink fieldCode="DE" term="%22Cultural+values%22">Cultural values</searchLink><br /><searchLink fieldCode="DE" term="%22Algorithmic+bias%22">Algorithmic bias</searchLink><br /><searchLink fieldCode="DE" term="%22Natural+language+processing%22">Natural language processing</searchLink><br /><searchLink fieldCode="DE" term="%22Social+values%22">Social values</searchLink><br /><searchLink fieldCode="DE" term="%22Language+%26+languages%22">Language & languages</searchLink><br /><searchLink fieldCode="DE" term="%22Language+models%22">Language models</searchLink><br /><searchLink fieldCode="DE" term="%22Social+attitudes%22">Social attitudes</searchLink> – Name: SubjectGeographic Label: Geographic Terms Group: Su Data: <searchLink fieldCode="DE" term="%22Netherlands%22">Netherlands</searchLink><br /><searchLink fieldCode="DE" term="%22United+States%22">United States</searchLink><br /><searchLink fieldCode="DE" term="%22Japan%22">Japan</searchLink><br /><searchLink fieldCode="DE" term="%22Germany%22">Germany</searchLink> – Name: Abstract Label: Abstract Group: Ab Data: Large language models (LLMs) are rapidly being adopted by users across the globe, who interact with them in a diverse range of languages. At the same time, there are well-documented imbalances in the training data and optimization objectives of this technology, raising doubts as to whether LLMs can accurately represent the cultural diversity of their broad user base. In this study, we look at LLMs and cultural values in particular, and examine how prompt language and cultural framing influence model responses and their alignment with human values in different countries. We do so by probing 10 LLMs with 63 items from the Hofstede Values Survey Module and World Values Survey, translated into 11 languages, and formulated as prompts with and without different explicit cultural perspectives. Our study confirms that both prompt language and cultural perspective produce variation in LLM outputs, but with an important caveat: While targeted prompting can, to a certain extent, steer LLM responses in the direction of the predominant values of the corresponding countries, it does not overcome the models' systematic bias toward the values associated with a restricted set of countries in our dataset: the Netherlands, Germany, the United States, and Japan. All tested models, regardless of their origin, exhibit remarkably similar patterns: They produce fairly neutral responses on most topics, with selective progressive stances on issues such as social tolerance. Alignment with cultural values of human respondents is improved more with an explicit cultural perspective than with a targeted prompt language. Unexpectedly, combining both approaches is no more effective than cultural framing with an English prompt. These findings reveal that LLMs occupy an uncomfortable middle ground: They are responsive enough to changes in prompts to produce variation, but they are also too firmly anchored to specific cultural defaults to adequately represent cultural diversity. [ABSTRACT FROM AUTHOR] – Name: AbstractSuppliedCopyright Label: Group: Ab Data: <i>Copyright of Computational Linguistics is the property of MIT Press and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract.</i> (Copyright applies to all Abstracts.) |
| PLink | https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=egs&AN=194945038 |
| RecordInfo | BibRecord: BibEntity: Identifiers: – Type: doi Value: 10.1162/COLI.a.583 Languages: – Code: eng Text: English PhysicalDescription: Pagination: PageCount: 88 StartPage: 407 Subjects: – SubjectFull: Cultural values Type: general – SubjectFull: Algorithmic bias Type: general – SubjectFull: Natural language processing Type: general – SubjectFull: Social values Type: general – SubjectFull: Language & languages Type: general – SubjectFull: Language models Type: general – SubjectFull: Social attitudes Type: general – SubjectFull: Netherlands Type: general – SubjectFull: United States Type: general – SubjectFull: Japan Type: general – SubjectFull: Germany Type: general Titles: – TitleFull: LLMs and Cultural Values: The Impact of Prompt Language and Explicit Cultural Framing. Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Bulté, Bram – PersonEntity: Name: NameFull: Terryn, Ayla Rigouts IsPartOfRelationships: – BibEntity: Dates: – D: 01 M: 06 Text: Jun2026 Type: published Y: 2026 Identifiers: – Type: issn-print Value: 08912017 Numbering: – Type: volume Value: 52 – Type: issue Value: 2 Titles: – TitleFull: Computational Linguistics Type: main |
| ResultId | 1 |
