Contribution to the handbook
Supporting · Assessed 2026-10-27
The evidence offers limited background on screening AI-generated discussion questions for safety, ethics, clarity, and breadth of views. It does not establish review effectiveness or specify a governance mechanism, so its contribution is supporting rather than central.
Does this challenge the goal?
In-scope gap
The question concerns how discussion prompts were generated and reviewed, a possible safeguard detail within deliberative system design. It does not contest the definition’s goal.
The supplied accounts differ in reported generation details, and neither establishes how effective the review was.
Why this category?
- Safeguard background
- Limits in review criteria and effectiveness
The two supplied interpretations report different counts and review descriptions; the packet does not resolve the discrepancy or establish review outcomes.
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 goalNew evidence is being assessed. Any answer below is provisional; parts whose evidence changed or was withdrawn are withheld.
Describe how the discussion questions were generated and what criteria were used to review them.
What we can say so far
The authors used Chinchilla 70B to generate 10,000 discussion questions from 175 author-written examples. They reviewed questions for safety, ethics, clarity, and whether they could support a range of views. Another account says screening aimed to reduce the risk of offensive comments and reports about 5,500 distinct questions; the supplied evidence does not explain how that figure relates to the 10,000 generated questions.
What remains open
- The evidence does not reconcile the report of 10,000 generated questions with the report of about 5,500 distinct questions.
- The review outcomes and how reviewers resolved disagreements are not described.
Conditions and limits
- This is a report of the study's methods based on the readable article text; supplementary materials are not included.
- The evidence does not establish that every question was free of risk or bias.
Objections to consider
- Its reported count of about 5,500 distinct questions differs from the 10,000 generated-question count, and the supplied evidence does not explain the relationship.
Evidence and related ideas
Supporting evidence, qualifications and context play different roles in this answer.
supports
Questions were generated and reviewed The authors used Chinchilla 70B to generate 10,000 discussion questions from 175 author-written examples. They reviewed the questions for safety, ethics, clarity, and whether they could support a range of views.
Provides the generation model, number of author-written examples, generated-question count, and review criteria.
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) ↗Read source discussion: AI can help humans find common ground in democratic deliberation →
qualifies
Generating and screening questions Researchers prompted a language model with a small set of examples to generate candidate discussion questions. They reviewed the questions to reduce the risk of offensive comments and assembled about 5,500 distinct questions.
Adds the screening aim and reports about 5,500 distinct questions, without explaining how that count relates to the generated-question count.
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) ↗Read source discussion: AI can help humans find common ground in democratic deliberation →