Cultural bias and cultural alignment of large language models
PNAS Nexus · 02.09.2024 · 3 (9)
DOI: 10.1093/pnasnexus/pgae346 ↗Kurz gefasst
Fachartikel in der Zeitschrift PNAS Nexus, online erschienen am 17.09.2024. Verfasst von Yan Tao und 3 weiteren Autor:innen. Beteiligte Einrichtung: Department of Information Science, Cornell University, Ithaca, NY 14853, USA. Die Arbeit wurde bisher 409-mal zitiert (Crossref) und nennt 50 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0). Gefördert von: Jacobs Foundation Research Fellowship and Digital Futures.
Abstract (Originalsprache)
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 later 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.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Information Science, Cornell University, Ithaca, NY 14853, USA.
- Department of Human Centered Technology, KTH Royal Institute of Technology, Stockholm 10044, Sweden.
- Graduate School of Education, University of Pennsylvania, Philadelphia, PA 19104, USA.
ORCID-Profile: Yan Tao, Olga Viberg, Ryan S Baker, René F Kizilcec
Förderung
- Jacobs Foundation Research Fellowship and Digital Futures
Zitieren
APA
Tao, Y., Viberg, O., Baker, R. S., & Kizilcec, R. F. (2024). Cultural bias and cultural alignment of large language models. PNAS Nexus, 3(9). https://doi.org/10.1093/pnasnexus/pgae346
Vancouver
Tao Y, Viberg O, Baker RS, Kizilcec RF. Cultural bias and cultural alignment of large language models. PNAS Nexus. 2024;3(9). doi:10.1093/pnasnexus/pgae346
BibTeX
@article{Tao2024,
author = {Tao, Yan and Viberg, Olga and Baker, Ryan S and Kizilcec, René F},
title = {Cultural bias and cultural alignment of large language models},
journal = {PNAS Nexus},
year = {2024},
volume = {3},
number = {9},
doi = {10.1093/pnasnexus/pgae346}
}
EUVIMED