Contribution to the handbook
Supporting · Assessed 2026-10-30
The comparisons provide useful background for interpreting evidence about AI-assisted group statements, rather than directly specifying a governance system. Participant and outside-judge measures are distinct, and the human comparison used a different critique procedure.
Does this challenge the goal?
None identified
The question examines a study’s comparisons and uncertainty reporting; it does not contest the exploration’s goal or desired outcome.
Its usefulness depends on whether the handbook draws on this study.
Why this category?
- Supporting evidence interpretation
- Limits of a comparison
The figure caption reports bootstrapped 95% confidence intervals but does not provide detailed outcomes for every measure. The records are interpretations of the same study, not independent corroboration.
This is an AI assessment of contribution to the goal, separate from whether the question has an answer.
Making and interpreting laws →
Tests, stop conditions and unresolved work →
Read the exploration goalIdentify which kinds of statements and judgments were compared, including participant and outside-judge measures where applicable.
What we can say so far
The study compared AI-generated group statements with statements written by trained, incentivized human mediators. Participants ranked statements by agreement and rated argument quality, endorsement and perceived bias. Separately, 155 outside raters judged features including clarity, informativeness, logic, representation of the majority, the writer’s own opinion, fair summary, inclusion of minority views and polarization. The researchers also compared initial statements with versions revised after participant critiques, and compared statements from an AI that had seen critiques with those from one that had not. They analyzed how individual opinions and initial and revised group statements related on a question-specific position scale, including estimated weight given to minority views. The human mediators did not have the AI process’s critique phase, so the procedures were not identical. Statement length also affected ratings and rankings. Outside raters found no statistically significant difference on fair summary, minority inclusion or polarization; endorsement did not necessarily mean agreement with every part of a statement.
What remains open
- The supplied descriptions do not specify the detailed wording or scales for every rating measure.
Conditions and limits
- The human-mediator comparison did not use an identical critique procedure.
- Statement length affected ratings and rankings, although the reported endorsement difference persisted after controlling for length.
- The text-position and minority-weight analyses depend on chosen representations and modeling assumptions.
- The descriptions are interpretations of the same article, not independent corroboration; supplementary materials were not supplied.
Objections to consider
- Human mediators lacked a critique phase, so the comparison does not isolate mediator type under matched procedures.
- Length affected judgments; adjustment preserved the reported endorsement difference but does not eliminate its influence on other measures.
- Outside raters found no statistical difference on fair summary, minority inclusion or polarization; the answer preserves these specific results rather than claiming uniformly higher ratings.
- A participant’s endorsement cannot establish agreement with every part of the statement.
- The named tasks used different procedures, whose consequences are not stated.
Evidence and related ideas
Supporting evidence, qualifications and context play different roles in this answer.
context
The study tested AI support for group discussion The study asked whether AI could help people with different views produce a group statement they would accept together.
States the study aim.
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 →
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.
Describes drafting and revision.
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 →
context
The study examined UK groups The experiments involved 5,734 participants from the United Kingdom and examined common ground, changes in participants’ views, how statements handled minority and majority views, and possible bias. The researchers also tested the approach in a virtual citizens’ assembly with a sample described as demographically representative of UK residents.
Describes samples and research questions.
AI can help humans find common ground in democratic deliberation
Read full source ↗Describe how uncertainty was displayed for the reported comparisons.
What we can say so far
The study’s figures displayed uncertainty with error bars showing bootstrapped 95% confidence intervals. In the agreement figure, fitted regression trend lines were also shown with standard errors. The reported average statement lengths were accompanied by intervals: 125.9 words for AI statements (95% CI 121.4–130.1) and 105 words for human statements (100.2–108.5).
What remains open
No material gap is recorded within this part’s current scope.
Conditions and limits
- The agreement figure included only rounds in which all five group members were present both before and after discussion.
- The evidence is an AI interpretation of the shared article text, not independent verification; supplementary materials were not included.
Evidence and related ideas
Supporting evidence, qualifications and context play different roles in this answer.
supports
AI statements were longer on average AI-written statements averaged 125.9 words (95% confidence interval 121.4–130.1), compared with 105 words for human-written statements (100.2–108.5). Length affected ratings and rankings. However, participants still preferred AI statements on endorsement ratings after the researchers controlled for length.
Provides the reported statement-length averages and their intervals as an example.
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 →
supports
The figure compared statements in several ways The figure caption describes several ways the study compared AI-generated and human-written group statements, including rankings that favored one over the other. It also compares revised statements that included critiques with initial statements that left them out. The figure’s error bars show bootstrapped 95% confidence intervals.
Reports that the statement-comparison figure used bootstrapped 95% confidence-interval error bars.