Observed arrival · 2026-08-29
AskAI: Can Silicon Samples Be Trusted?
A research project testing whether AI-simulated survey respondents can produce trustworthy public-opinion data.
Field notes
The project models a survey respondent by prompting a large language model to role-play a member of a population, then turns established survey methodology and psychometrics on the resulting data. Its proposed framework, “siliconometrics,” is intended to distinguish contexts where simulated samples work from those where they fail. The homepage links to papers, a blog, and a podcast, while identifying an EXCELLENCE_25 ERC preparatory grant and an institutional base at the ELTE Centre for Social Sciences.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue