Reproduce “Using AI Predictions to Augment Rather Than Replace Surveys” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Lai Wei, Yifei Zhu
The authors test prediction-powered inference (PPI) for improving estimates from probability surveys by combining model predictions on large auxiliary files with survey responses that correct prediction errors. Across 83 GSS and NHIS outcomes, NHIS mean estimation shows a 12.2% effective-sample-size gain with historical supervised learning, versus 2.6% with GPT-5-mini. Gains depend on aligned populations, covariates, and appropriate weights.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.