Characterizing the ability of LLMs to recapitulate Americans' distributional responses to public opinion polling questions across political issues

Eric Gong and BKC Affiliates Nathan Sanders and Bruce Schneier propose more accurate and lower-cost method of using LLMs to augment human population studies in polling contexts. In a paper published on arXiv, the authors query an LLM (GPT 4o-mini) with questions for which human polling data has already been collected via the 2022 Cooperative Election Study.