GPT-4 predicts 70 social science experiments with r=.85 across 119,330 participants
RobbWiller · x · 2026-07-23
Main finding
LLM-derived effect estimates from simulated participants were highly correlated with observed effects across 70 preregistered social science experiments: r = .85 and adjusted r = .92.
What the paper did
- Built an archive of 70 nationally representative TESS studies plus replications
- Covered 469 effects and 119,330 participants
- Prompted GPT-4 with study materials and demographic profiles, then inferred treatment effects from simulated responses
Key takeaway
The correlations stayed high across a wide range of social-science domains, and the figure shows LLM predictions were broadly comparable to the average expert forecasts.
Related event: Nature Study: GPT-4 Can Predict Social Science Experiment Results(18 posts)→
More from Research
- Meta FAIR's Self-Direction turns agent failures into reusable LoRA skills, +13.8pp on Terminal-Bench — karen_ullrich · 2026-10-27
- DONIC hits 86.79% on bimanual dexterous tracking, crushing prior SOTA's 9.43% — liuziwei7 · 2026-10-27
- Why don't more LLMs ship 10M-token context? A researcher asks about attention tradeoffs — agarwl_ · 2026-10-27
- PSNR below 15 dB in papers likely means the metric was misused, researcher warns — kwangmoo_yi · 2026-10-27
- Ortet's first health eval: 34 open-weight models, 50 tasks, generic rankings fall short — thegautamkamath · 2026-10-27
- Scaling-fit discrepancy: output-head FLOPs are ~65% of compute at the smallest rung, ~3% at 535B — dlwh · 2026-10-27