Not All Countries Celebrate Thanksgiving: On the Cultural Dominance in Large Language Models

TL;DR

This paper reveals that large language models tend to favor English and Western cultural responses due to training data biases, and proposes evaluation benchmarks and mitigation strategies to address this cultural dominance.

Contribution

It introduces a systematic benchmark for evaluating cultural bias in LLMs and demonstrates effective mitigation methods like diverse data pretraining and culture-aware prompting.

Findings

GPT-4 exhibits the strongest cultural dominance bias.

Pretraining on diverse data reduces cultural bias.

Culture-aware prompting significantly mitigates bias.

Abstract

This paper identifies a cultural dominance issue within large language models (LLMs) due to the predominant use of English data in model training (e.g., ChatGPT). LLMs often provide inappropriate English-culture-related answers that are not relevant to the expected culture when users ask in non-English languages. To systematically evaluate the cultural dominance issue, we build a benchmark of concrete (e.g., holidays and songs) and abstract (e.g., values and opinions) cultural objects. Empirical results show that the representative GPT models suffer from the culture dominance problem, where GPT-4 is the most affected while text-davinci-003 suffers the least from this problem. Our study emphasizes the need to critically examine cultural dominance and ethical consideration in their development and deployment. We show that two straightforward methods in model development (i.e., pretraining on more diverse data) and deployment (e.g., culture-aware prompting) can significantly mitigate the cultural dominance issue in LLMs.