Nature review places LLMs in four social-science roles

A new review in Nature Human Behaviour sorts how large language models are being used in computational social science into four research roles: as subjects that simulate human behavior, as assistants that automate analysis, as collaborators that generate hypotheses, and as evaluators that audit and improve research quality.

The review, led by Fengli Xu and James Evans with colleagues at Tsinghua University, HKUST and Microsoft Research Asia, was published on 26 October 2026. It documents advances across economics, psychology, political science, sociology and anthropology.

The authors are blunt about the limits. LLMs "struggle with causal inference and raise urgent ethical concerns about privacy and misinformation," they write, and their ability to reproduce social complexity "remains uneven and unproven." When LLMs stand in for human participants, the review notes, they can "harmfully misportray and flatten identity groups."

Two researchers we track in our Who's Who directory circulated the paper on its first day online.