Are Generative Language Models Multicultural? A Study on Hausa Culture and Emotions using ChatGPT

TL;DR

This study evaluates how well ChatGPT captures Hausa culture and emotions by comparing its responses to native speakers, revealing partial alignment but notable gaps and biases in representing low-resource language contexts.

Contribution

It introduces a methodology for assessing cultural and emotional representation of LLMs in low-resource languages, specifically Hausa, highlighting existing gaps and biases.

Findings

ChatGPT shows some similarity to native responses

Gaps and biases exist in ChatGPT's cultural knowledge

Evaluation methodology for low-resource languages is proposed

Abstract

Large Language Models (LLMs), such as ChatGPT, are widely used to generate content for various purposes and audiences. However, these models may not reflect the cultural and emotional diversity of their users, especially for low-resource languages. In this paper, we investigate how ChatGPT represents Hausa's culture and emotions. We compare responses generated by ChatGPT with those provided by native Hausa speakers on 37 culturally relevant questions. We conducted experiments using emotion analysis and applied two similarity metrics to measure the alignment between human and ChatGPT responses. We also collected human participants ratings and feedback on ChatGPT responses. Our results show that ChatGPT has some level of similarity to human responses, but also exhibits some gaps and biases in its knowledge and awareness of the Hausa culture and emotions. We discuss the implications and limitations of our methodology and analysis and suggest ways to improve the performance and evaluation of LLMs for low-resource languages.