Contribution to the handbook
Supporting · Assessed 2026-10-27
The evidence identifies component ablations, tests on new topics, and preregistration of sample sizes, exclusions, and primary analyses. This helps characterize the study’s testing approach, but the packet does not provide all results or a complete list of the seven evaluations, so it is supporting rather than a basis for a design recommendation.
Does this challenge the goal?
In-scope gap
The question concerns what evaluations were conducted and what they assessed, which is relevant to testing possible system components within the existing goal; it does not contest that goal.
The supplied passage is incomplete, so the complete evaluation list and some results are not established here.
Why this category?
- Provides context on component testing and evaluation planning
- The evidence explicitly notes that the list is incomplete and some findings are absent
- Useful for test design background but does not establish governance-scale effectiveness
The records are interpretations of a shared article, supplementary materials are absent, and the supplied evaluation passage cuts off before completing its list.
This is an AI assessment of contribution to the goal, separate from whether the question has an answer.
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 Any answer below is provisional; parts whose evidence changed or was withdrawn are withheld.
Identify the kinds of evaluations the passage lists and state what the main task and ablation experiments assessed; do not fill gaps in an incomplete list by inference.
What we can say so far
The authors report seven evaluation experiments. The available descriptions name the main task and ablations, tests of the critique phase, mediation’s effect on belief change, a human-mediator experiment, a virtual UK citizens’ assembly, an unmediated opinion-exposure comparison, and checks on discussion topics not used in training. The main task and ablations assessed the importance of different Habermas Machine components; the additional system checks specifically mention the reward model and supervised fine-tuning. The passage’s list cuts off, so these descriptions should not be treated as a reconstructed seven-item list or as a full account of the human-mediator experiment.
What remains open
- The truncated list does not establish a complete, distinct seven-item inventory or finish describing the human-mediator experiment.
Conditions and limits
- This reports evaluation kinds and stated aims, not evidence that an evaluation succeeded.
- The assembly, opinion-exposure comparison, and new-topic checks are identified in other supplied article summaries, not in the truncated list itself.
Evidence and related ideas
Supporting evidence, qualifications and context play different roles in this answer.
supports
The study tested several parts The authors say they ran seven evaluation experiments. The main task and ablation experiments assessed the importance of different Habermas Machine components. The visible list also names tests of the critique phase, mediation’s effect on belief change, and a human-mediator experiment. The passage cuts off before describing that last experiment or listing all seven.
States the reported count, visible evaluation labels, component-assessment purpose, and truncation.
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
Agreement also increased in the assembly The researchers tested the process in a virtual citizens’ assembly with 200 UK participants recruited to reflect the population across age, gender identity, ethnicity, socioeconomic status, and region. Assembly groups used the same deliberation procedure as the main-task cohorts. Support for group opinions was similarly high, and group agreement increased significantly from before to after deliberation, as it did in the crowd-sourced cohorts.