Skip to content

Advisors that disagree never decide by count: a conclusion is refused while a disagreement on its assessment is unsettled, a resolution needs evidence, and every answer received must be weighed

When advice diverges, the divergence is a record — the positions, the assumptions they

declared statusdeclared status: guaranteed Declared deterministic and blocking, implemented, and naming the behavioural test that is to prove it. Whether a recorded run supports it is the evidence verdict, shown beside it.

recorded evidencerecorded evidence: not run Declared guaranteed, and not supported by a recorded run: the test is named, and no passing run of the current inputs is recorded. how evidence is recorded

the case lets the minority position win once an experiment settles the disagreement.

What it means

When advice diverges, the divergence is a record — the positions, the assumptions they divide on, the question that decides — and the decision waits for evidence. Two advisors agreeing is evidence about agreement; it settles nothing on its own.

How it works

The writer (apps/majordomus-cli/src/reasoning/store.rs) refuses a conclusion on an assessment with an unresolved disagreement, refuses a resolution that cites no evidence, and refuses a conclusion that leaves out an answer its assessment received (it may reject the answer in rejected, not ignore it). The review count is computed from the cited consultations; there is no field to author it.

How to see it

majordomus reasoning record --file disagreement.json
majordomus reasoning explain <conclusion-id>   # the experiment is in the chain

test/cases/732_disagreement_is_settled_by_evidence.sh consults two scripted advisors that disagree, is refused while two of three positions agree, and concludes the minority's position after an experiment.

What it does not cover

It cannot judge whether an experiment was a good one. It makes the experiment and its evidence part of the record, where a reviewer can judge it.

Why it exists

Majority vote over models is the failure mode of naive multi-model review: it rewards correlated errors. Making the count unable to decide anything is the repair (ADR 0098).

Detail rendered from docs/claims/advice-is-weighed-not-counted.md.

From the statement to the proof

  1. 1 · the public statement

    Advisors that disagree never decide by count: a conclusion is refused while a disagreement on its assessment is unsettled, a resolution needs evidence, and every answer received must be weighed

    Made on Reasoning.

  2. 2 · the canonical claim

    advice-is-weighed-not-counteddeclared status: guaranteed in docs/CLAIMS.yaml

    Defined in read it on this site · docs/REASONING.md

  3. 3 · the implementation

    apps/majordomus-cli/src/reasoning/store.rs

  4. 4 · the test and the scenarios

    Test: test/cases/732_disagreement_is_settled_by_evidence.sh

  5. 5 · the recorded runs

    recorded evidence: not run how evidence is recorded

Verify it yourself

The test runs in a disposable temporary repository and asserts the behaviour, not a string in the source.

from a clone of the repository
bash test/run.sh 732_disagreement_is_settled_by_evidence

Searches the titles and descriptions of every page this site publishes.