Not All Countries Celebrate Thanksgiving: On the Cultural Dominance in Large Language Models
This addresses cultural bias in LLMs for global users, though it is incremental as it builds on existing bias research with specific cultural benchmarks.
The paper identifies cultural dominance in large language models due to English-centric training data, showing that models like GPT-4 provide inappropriate English-culture answers for non-English queries. It demonstrates that methods like diverse pretraining and culture-aware prompting can significantly mitigate this issue.
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.