Large Language Models Reproduce Racial Stereotypes When Used for Text Annotation

Petter Törnberg

Large language models (LLMs) are increasingly used for automated text annotation in tasks ranging from academic research to content moderation and hiring. Across 19 LLMs and two ex…