Contribution to the handbook
Priority · Assessed 2026-10-28
The additional cohorts and tests address whether system components matter and whether results carry over to topics omitted from language-model fine-tuning. These are directly relevant test dimensions for a governance system that must be evaluated beyond its initial comparison. The available records describe test aims and the held-out-topic setup, but do not provide those tests’ results; they therefore cannot support a claim of successful transfer or component effectiveness.
Does this challenge the goal?
None identified
The question asks what further evaluations the researchers conducted; it does not contest the handbook’s goal or scope.
The packet reports test aims and methods but does not provide the results needed to judge the system’s performance.
Why this category?
- Material testing dimensions: component contribution and transfer to unfamiliar topics
- Explicit distinction between evaluation aims and unavailable results
The supplied records do not report the additional tests’ findings, provide the full list of evaluations, or establish transfer beyond the tested study conditions.
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 goalIdentify the additional cohorts and what they tested about system components and transfer to topics not used in training, distinguishing test aims from reported results.
What we can say so far
The authors collected three additional cohorts to examine the importance of system components, including the reward model and supervised fine-tuning, and to test transfer to topics not used in training. For the unfamiliar-topic test, 12 topics were held out of the language-model fine-tuning data. These are reported test aims and methods; the supplied passages do not report the outcomes of these tests.
What remains open
- The outcomes of the component and held-out-topic tests are not available in the supplied evidence.
- The scope of the 12 held-out topics is not described in the supplied evidence.
Conditions and limits
- This is based on AI interpretations of the same Science article text, not independent evidence.
- The supplied article copy excludes supplementary materials.
- The passages do not provide the results of the additional tests.
Evidence and related ideas
Supporting evidence, qualifications and context play different roles in this answer.
supports
Further cohorts tested system parts and new topics The researchers collected three additional cohorts to test the importance of parts of the AI system, including its reward model and supervised fine-tuning, and to check whether results carried over to discussion topics not used in training.
Identifies three additional cohorts and their aims: testing system parts, including the reward model and supervised fine-tuning, and checking transfer to topics not used in training; it also notes that results are not supplied.
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 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.