LLMs show Western moral bias in study

Large language models including OpenAI's GPT-3.5, GPT-4, and GPT-4o systematically align with Western moral values, overestimating concerns like care while underestimating values such as purity for non-Western nations. Published in the Proceedings of the National Academy of Sciences, the study compared AI estimates against over 90,000 human participants across 48 nations.