Contribution to the handbook
Supporting · Assessed 2026-10-28
The evidence specifies candidate generation, personalized predicted rankings, and how those rankings were combined to select a winner. That is useful mechanism background, but the supplied records do not establish prediction accuracy, fairness of the winner, or suitability for public decisions.
Does this challenge the goal?
In-scope gap
The question concerns how a group statement was selected, a mechanism relevant to exploring deliberative design. It does not challenge the exploration’s goals or require adopting this selection method.
The evidence describes a study procedure, not a validated decision rule for public governance.
Why this category?
- system mechanism background
- limits of evidence
These records derive from the same article. They describe the study’s procedure, not the accuracy of predicted rankings, fairness of the winner, or suitability for a government decision.
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 goalExplain candidate generation, predicted individual rankings, and how those predictions were combined to select a winner; do not imply the predictions were necessarily accurate.
What we can say so far
The helper generated 16 candidate statements. For each participant, a personalized reward model used that person’s opinion and critique to predict a ranking of the candidates. The system combined those predicted rankings using the Schulze voting method and returned the winning statement. This describes the reported selection mechanism; it does not establish that the predictions were accurate or that the winner represented everyone fairly.
What remains open
- The supplied evidence does not report the accuracy of the predicted rankings or establish that the selected statement represented every participant fairly; those outcomes are not needed to describe the mechanism.
Conditions and limits
- This is the method reported in the supplied copy of the Science article; supplementary materials are not included.
- The account describes a simulated election among generated candidates, not proof that the method is best for public decisions or that its predictions were accurate.
- The supplied evidence does not independently verify the article’s account.
Objections to consider
- The evidence says the reward model was intended to predict agreement but provides no validation of its accuracy. The answer reports predictions as predictions and explicitly does not imply they were accurate.
- This excerpt is incomplete about implementation, but the detailed account of candidate generation and aggregation is supplied separately.
Evidence and related ideas
Supporting evidence, qualifications and context play different roles in this answer.
context
The AI revised statements using feedback The AI helper drafted a group statement from participants’ opinions, then revised it using their critiques. The design drew on Jürgen Habermas’s theory of communicative action; across successive rounds, researchers used supervised fine-tuning and reward modeling to improve the system, with group approval as the stated goal.
Provides general context that the helper drafted and revised group statements, but not the detailed selection procedure.
AI can help humans find common ground in democratic deliberation
Read full source: AI can help humans find common ground in democratic deliberation (opens in a new tab) ↗