Cultural Bias and Cultural Alignment of Large Language Models

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Tao, Yan, Viberg, Olga, Baker, Ryan S., Kizilcec, Rene F.
Natura: Preprint
Pubblicazione: 2023
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916400281419776
author Tao, Yan
Viberg, Olga
Baker, Ryan S.
Kizilcec, Rene F.
author_facet Tao, Yan
Viberg, Olga
Baker, Ryan S.
Kizilcec, Rene F.
contents Culture fundamentally shapes people's reasoning, behavior, and communication. As people increasingly use generative artificial intelligence (AI) to expedite and automate personal and professional tasks, cultural values embedded in AI models may bias people's authentic expression and contribute to the dominance of certain cultures. We conduct a disaggregated evaluation of cultural bias for five widely used large language models (OpenAI's GPT-4o/4-turbo/4/3.5-turbo/3) by comparing the models' responses to nationally representative survey data. All models exhibit cultural values resembling English-speaking and Protestant European countries. We test cultural prompting as a control strategy to increase cultural alignment for each country/territory. For recent models (GPT-4, 4-turbo, 4o), this improves the cultural alignment of the models' output for 71-81% of countries and territories. We suggest using cultural prompting and ongoing evaluation to reduce cultural bias in the output of generative AI.
format Preprint
id arxiv_https___arxiv_org_abs_2311_14096
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Cultural Bias and Cultural Alignment of Large Language Models
Tao, Yan
Viberg, Olga
Baker, Ryan S.
Kizilcec, Rene F.
Computation and Language
Artificial Intelligence
Culture fundamentally shapes people's reasoning, behavior, and communication. As people increasingly use generative artificial intelligence (AI) to expedite and automate personal and professional tasks, cultural values embedded in AI models may bias people's authentic expression and contribute to the dominance of certain cultures. We conduct a disaggregated evaluation of cultural bias for five widely used large language models (OpenAI's GPT-4o/4-turbo/4/3.5-turbo/3) by comparing the models' responses to nationally representative survey data. All models exhibit cultural values resembling English-speaking and Protestant European countries. We test cultural prompting as a control strategy to increase cultural alignment for each country/territory. For recent models (GPT-4, 4-turbo, 4o), this improves the cultural alignment of the models' output for 71-81% of countries and territories. We suggest using cultural prompting and ongoing evaluation to reduce cultural bias in the output of generative AI.
title Cultural Bias and Cultural Alignment of Large Language Models
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2311.14096