Contribution to the handbook
Priority · Assessed 2026-10-28
The evidence reports higher accuracy for liberal than conservative and centrist participants, and for college-educated than non-college-educated participants in the study. If predictions informed whose preferences were represented, these differences could matter materially. The records do not establish real-world overrepresentation or broader effects.
Does this challenge the goal?
In-scope gap
Testing prediction accuracy across political-view and education groups fits the existing citizen-serving goal and does not contest it.
The findings concern a specific participant group and set of models, not performance in a governance system.
Why this category?
- Identifies observed accuracy differences relevant to representation of citizens’ preferences.
- Provides a concrete group-testing concern while preserving limits on the findings’ reach and consequences.
The evidence comes from one study and shared source material. It does not establish the size or cause of the gaps, or whether they would persist in other populations or decision processes.
This is an AI assessment of contribution to the goal, separate from whether the question has an answer.
Authority, accountability and resilience →
Tests, stop conditions and unresolved work →
Read the exploration goalState whether and in what direction accuracy differed by political view in the study.
What we can say so far
Yes. In this study, all six models were reported to predict choices more accurately for liberal participants than for conservative and centrist participants.
What remains open
- The size of the accuracy differences is not provided.
Conditions and limits
- This is a reported result for the participants and models in this study; it does not establish the same pattern in other populations or with other models.
- The supplied evidence does not explain the cause of the differences or quantify their size.
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.
Describes the volunteer policy-choice data and its limits; it does not report accuracy differences by political view.
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 →
context
Some groups may be missing Biases and other limits may make language models worse at representing some political views, especially views of groups missing from training data.
Notes a general limitation concerning representation of groups; it does not report the study's accuracy differences by political view.
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) ↗State whether and in what direction accuracy differed by education in the study.
What we can say so far
Yes. The authors report higher prediction accuracy for college-educated than non-college-educated participants. They report statistically significant average-accuracy differences between these groups for all six models. These findings concern the participants and models in this study.
What remains open
- The supplied evidence does not give the size of the accuracy difference.
Conditions and limits
- The finding is limited to the participants and models examined in the study; it does not establish that the pattern generalizes to other populations or models.
- The supplied summaries are AI interpretations of the study text; the figures were not independently interpreted.
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.
Describes the study's volunteer policy-choice data and limits on representativeness, not education-related accuracy results.
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 →
context
Some groups may be missing Biases and other limits may make language models worse at representing some political views, especially views of groups missing from training data.
Gives a general caution about representing groups, without reporting education-specific accuracy results.
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) ↗