Nature review casts LLMs in four roles for social science

Fengli Xu, James Evans and colleagues at Tsinghua University, HKUST, Microsoft Research Asia and the University of Chicago published a review on 26 October 2026 in Nature Human Behaviour arguing that large language models now occupy four distinct roles in computational social science: subjects that simulate human behaviour, assistants that automate analysis, collaborators that generate hypotheses, and evaluators that audit and improve research quality.

The strongest claims sit in the 'subject' column. When prompted with demographic or ideological profiles, the authors write, models 'reproduce survey distributions, economic games and polarization dynamics.' AI interviewers, they argue, deliver protocols 'with quality comparable to human interviews.' Models that code qualitative data act as 'alien readers,' surfacing patterns a human coder would miss.

The hedges are blunt. The authors say LLMs 'struggle with causal inference and raise urgent ethical concerns about privacy and misinformation,' and that their 'ability to reproduce social complexity remains uneven and unproven.' They frame the field with a 'recursive challenge': the tools reshaping society are also the tools being used to study that reshaping.

The review spans sociology, political science, psychology, anthropology and economics. Two researchers from our tracker posted the source link within days of publication, a signal the paper is moving through AI-research feeds beyond its home discipline.