Characterizing the ability of LLMs to recapitulate Americans’ distributional responses to public opinion polling questions across political issues
Eric Gong [email protected], Nathan E. Sanders [email protected], Bruce Schneier [email protected] · 2026
Read Characterizing the ability of LLMs to recapitulate Americans’ distributional responses to public opinion polling questions across political issues by Eric Gong [email protected], Nathan E. Sanders [email protected], Bruce Schneier [email protected] (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-20229.txt.
Description
Gong, Sanders, and Schneier test if language models can predict how Americans answer opinion polls, as a distribution. Polling grows costly and suffers rising non response and thin demographic coverage. Their framework prompts a model once to output a full response distribution. They score it against the large Cooperative Election Study survey. The distribution method beat repeated single respondent simulation on accuracy. It also cost far less to run. Performance varied in systematic ways across demographics and questions.
Catalog record
| CREATOR | Eric Gong [email protected], Nathan E. Sanders [email protected], Bruce Schneier [email protected] |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 54777 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.20229