SCIENCE UNDER AUTOMATION
Simulated preferences across 74 countries
Twenty LLM configurations were compared with 67,201 matched respondents across 74 countries. Simulated preferences were unreliable proxies for representative cross-national surveys.
What this result supports
Partial pattern agreement does not replace recruitment, sampling or measurement of real participants.
Evidence record: Peer-reviewed evaluation of simulated survey responses. Checked: 2026-10-21.
Where this evidence is used
Social & behavioural science
- Recruit and measure an appropriate human population: Simulated respondents are not reliable replacements for representative cross-national survey measurements.
- Validate interpretation in the relevant social context: Cross-country errors and prompt sensitivity require contextual scrutiny.