Nature review slots LLMs into four social-science roles

Large language models now play four distinct working roles inside computational social science, according to a review published 26 October 2026 in Nature Human Behaviour. The authors, led by Fengli Xu with co-authors including Chenyang Shao, Lin Chen and James Evans, sort recent work across sociology, political science, psychology, anthropology and economics into those roles rather than by topic.

The roles the review names are "subjects that simulate human behaviour, assistants that automate analysis, collaborators that generate hypotheses and evaluators that audit and improve research quality." In the subjects role, the authors write that when prompted with demographic or ideological profiles, models "reproduce survey distributions, economic games and polarization dynamics" — a finding they present as suggestive rather than settled.

The review is explicit about what is not yet working. "LLMs struggle with causal inference and raise urgent ethical concerns about privacy and misinformation, and their ability to reproduce social complexity remains uneven and unproven," the paper says. The authors frame the whole project recursively: "As artificial intelligence reshapes human society, understanding this transformation requires the very tools driving it."

Two researchers on our Who's Who radar surfaced the piece on the day it ran, which is roughly the response rate we see when a Nature family journal formalises a methods vocabulary the field has been using informally.