Contribution to the handbook
Priority · Assessed 2026-10-28
The question brings together statistical methods, measures of written positions and bias, participant variation, and limits on inference. Those details can affect whether an apparent improvement in deliberation is a reliable test result or an artifact of measurement. The packet also includes important qualifications: operational bias tests are limited, causal explanations remain uncertain, and transfer beyond the UK is unresolved.
Does this challenge the goal?
None identified
The question asks how the study analyzed its results; it does not challenge the handbook’s goal, beneficiary boundary, or assumptions.
The analyses may be informative for evaluation design, but the study is not itself an evaluation of a city or country governance system.
Why this category?
- Evaluation methods and uncertainty materially affect proposed governance tests
- Evidence identifies measurement limits, bias measures, and limits on generalization
The evidence is interpretation of one article and shares a source root. Some supplied records describe methods without results, and the study’s measures do not establish fairness or causal effects in other settings.
This is an AI assessment of contribution to the goal, separate from whether the question has an answer.
- How did the authors analyze participants’ ratings and other results?
- Analysis methods
- Text scores used a question-specific scale
- Measuring majority lean
- Estimating minority influence
- A bias test found no clear link
- Majority exposure did not predict the shift
- Similar results in another UK sample
- Wider relevance remains uncertain
- The assembly left out fact-finding stages
Authority, accountability and resilience →
Tests, stop conditions and unresolved work →
Read the exploration goalIdentify the reported methods for analyzing ratings and other results, including exclusion procedures, rating models and participant/group variation, significance and uncertainty, written-opinion positions, minority influence, and operational bias measures. Preserve the limits of each measure.
What we can say so far
The article reports that sample sizes, exclusion criteria and primary analyses were preregistered. It also says the analyses used the preregistered exclusion procedures, ordered-response models for rating-scale answers, two-sided significance tests, and the most detailed random-effects structure the available data allowed to account for participant and group variation. The supplied account does not give the exclusion rules themselves, model specifications or estimates. For uncertainty, the figures are described as showing bootstrapped 95% confidence intervals; one agreement figure also uses fitted regression lines with standard errors. For written opinions, the authors converted text into numerical representations and located opinions and group statements on question-specific agreement–disagreement axes. A separate account describes 768-number text vectors rotated onto such an axis and checked against participants’ earlier ratings; the association was not exact. Minority influence was operationalized by classifying participants according to which side of neutral their ratings fell on, fitting group-statement positions to combinations of members’ opinions, and summing estimated weights for minority opinions, including analyses at different levels of group division. The bias measures included the proportion of generated statements leaning toward the majority rather than minority side of the group median, and, for the assembly, a comparison of statement positions with the agree and disagree sides relative to that median. These are the authors’ reported operational measures, not comprehensive measures of all possible bias or influence. The supplied evidence does not provide the registered records, detailed exclusion rules, complete model outputs, or results and validation sufficient to audit these procedures fully. The source is an interpretation of the article, and supplementary materials are not included.
What remains open
- The exact preregistered exclusion criteria and how exclusions were applied are not supplied.
- Model coefficients, test results, and enough model detail to assess the random-effects structures are not supplied here.
- The supplied descriptions do not establish the completeness or broader validity of the text-position, minority-influence, or bias measures.
Conditions and limits
- The evidence is an AI interpretation of the article, not independent verification; the supplied readable article copy includes references but not supplementary materials.
- The source reports preregistration and analysis choices but does not supply the registration records, detailed exclusion rules, full model specifications, or outputs.