Living Worlds

Research · Populations · Oct 5, 2026

A voice for a question nobody has asked before

Not met (1 of 5 inside the noise)five fresh 2024 questions

The test at a glance

What it tests
Accuracy against reality: the split by party and age, given each question's 2024 national share.
Data
General Social Survey 2024 (Release 3a); calibrated on 22 development rows (11 other questions in 2022 and 2024).
Held out
The 2024 party-by-age shares of five spending questions.
Compared against
Every group given the national share.
Bar
Lower mean error than the baseline AND inside the sampling floor on at least 3 of 5 questions.
Result
6.2 against 8.2 points, but inside the floor on 1 of 5: not met.

Asked directly for each group's share and levelled to each question's 2024 national share, a frontier model calibrated on eleven other questions split five fresh 2024 questions (highways, big cities, drug addiction, alternative energy, science) by party and age with a mean error of 6.2 points, ahead of giving every group the national share (8.2) and far ahead of the same model answering as each person. It landed inside the survey's own noise on 1 of 5, where 3 were needed. Registered before the model was asked anything about these questions. The next version is being tested on questions held out before it was trained.

Protocol, result and scorer

Next

The trained voice, on questions held out before it was trained.

Populations

More in this area.