Overview
The AI Personality Study examines if and how large language models adapt their perceived “personality” depending on a user’s language and cultural background. It is the first large-scale evaluation of human preferences for chatbot characteristics across diverse languages and cultural groups. It compares the performance of a range of LLMs (such as ChatGPT, Claude, LLama, Qwen, DeepSeek, Falcon) across English, Chinese, and Arabic.
The study is a collaboration between metaLAB (at) Harvard and the Max Planck Institute for Human Development. It was funded by Harvard University’s William F. Milton Fund Grant for Generative AI.