Debate Erupts Over Whether GPT-4 Can Predict Survey Experiment Outcomes

A Nature study found GPT-4 simulations can predict treatment effects of 70 preregistered survey experiments, but researcher Gerard Sans pushed back, arguing LLMs are text samplers rather than models of human psychology, sparking academic debate.

2026-10-25 ~ 2026-10-26 · 2 related posts

• John Horton highlights Nature study: GPT-4 simulates survey experiments and predicts real treatment effects — johnjhorton · 2026-10-25
• Researcher pushes back on LLM-simulated surveys: AI is a text sampler, not a model of human psychology — gerardsans · 2026-10-26