Contribution to the handbook
Priority · Assessed 2026-10-28
The question brings a concrete evaluation method and reported findings to the important problem of whether model-augmented samples can estimate group preferences. The evidence also makes the limits of that evaluation visible: matching a study sample is not evidence of representing people outside it.
Does this challenge the goal?
In-scope gap
The question examines how group-preference prediction was evaluated, a testing and citizen-representation issue within the approved goal. It does not contest that goal.
The evidence concerns one study’s participant group and does not establish transfer to a broader public or real governance decisions.
Why this category?
- Testing a preference-estimation method relevant to evaluating citizen representation
- Identifying the limits of the study’s comparison and reported results
The results are interpretations of one shared source, not independent replications. Some supplied passages describe the comparison without giving its outcome, and the reported improvement does not establish population representativeness or practical suitability.
This is an AI assessment of contribution to the goal, separate from whether the question has an answer.
Making and interpreting laws →
Authority, accountability and resilience →
Tests, stop conditions and unresolved work →
Read the exploration goalDescribe how ordinary and model-augmented samples were compared with the full group’s preferences.
What we can say so far
The researchers compare ordinary random samples with samples augmented by model-predicted preferences, assessing both against the aggregate preferences of the full 267-person study group. They summarize preferences by each proposal’s win rate—the share of recorded pairwise choices in which it was selected—and compare sample win rates with full-sample win rates using the squared Pearson correlation (R²). They treat 100% as identical win-rate patterns. The study describes the preference data as a participant-by-preference matrix and reports using cross-validation for its prediction assessment.
What remains open
- The supplied passages do not give the precise repeated-sampling or split procedure for every comparison.
Conditions and limits
- The benchmark is the study’s 267 participants, not the wider public.
- A match on proposal win-rate patterns does not establish that individual preferences or other features of the group are reproduced.
- This describes the researchers’ reported method; the supplied passages do not independently verify it.
Evidence and related ideas
Supporting evidence, qualifications and context play different roles in this answer.
context
The volunteer policy data The study uses volunteer choices between randomly paired policies from the two leading candidates’ 2022 Brazilian presidential programs. The choices came from an experiment involving collaborative construction of a government program.
Provides context about the volunteer policy-choice data used in the study.
Large Language Models (LLMs) as Agents for Augmented Democracy
Read full source: Large Language Models (LLMs) as Agents for Augmented Democracy (opens in a new tab) ↗Read source discussion: Large Language Models (LLMs) as Agents for Augmented Democracy →
supports
Estimates of group preferences Using cross-validation, the authors report that a probability-based sample augmented with model predictions estimated group policy preferences more accurately than the sample alone.
Reports cross-validation and the comparison of model-augmented samples with samples alone.