Initial scoring rubric · v0.2 · For calibration
Evaluating inquiry across an interaction sequence
Unit of assessment: one bounded episode, from the initial question through contributions, revisions, and subsequent work. Record the starting objective, the context available to each participant, and any adopted changes before assigning scores.
A well-supported challenge to the original premise can earn a high score. Multiple defensible research directions are possible; agreement with a preferred conclusion is not the criterion.
Inquiry development
- 0
- Silently substitutes a different objective or removes a distinction central to the inquiry.
- 1
- Preserves or clarifies the question and its relevant distinctions; adds no substantive new direction.
- 2
- Introduces a useful distinction, alternative explanation, or testable direction, and makes its relationship to the prior question explicit.
Warrant for change
- 0
- Treats an unsupported inference, model self-description, or conjecture as established evidence.
- 1
- Gives a relevant reason for a proposed change, but leaves its evidential basis, uncertainty, or alternatives partly unspecified.
- 2
- Connects the proposed change to identifiable evidence or reasoning, distinguishes observation from interpretation, and states what would support or challenge it.
Agency and attribution
- 0
- Misattributes a contribution or treats a proposal, inference, or disagreement as the person’s settled belief, endorsement, or authorization.
- 1
- Keeps proposals distinct from decisions, but leaves the origin of a material idea or the status of agreement unclear.
- 2
- Preserves who introduced and revised material ideas, records acceptance or unresolved disagreement, and respects the scope of delegated decisions.
Correction and repair persistence
- 0
- Later work contradicts an adopted correction or returns to a superseded objective without acknowledging the change.
- 1
- Retains the correction in its description but applies it only partly in later substantive work.
- 2
- The correction governs at least two later substantive decisions or outputs. If the record includes an interruption, the resumed work also preserves it.
Evidence required for each score
Record the dimension, score, turn IDs, speakers, exact supporting excerpts, and a brief rationale. Include relevant counterevidence. Preserve disagreements between reviewers for adjudication.
Not assessable: use N/A when the required evidence is absent. Persistence needs an adopted correction and at least two later opportunities to apply it. Claim recovery after interruption only when that transition is in the record.
Report the four dimensions separately. A productive new idea can coexist with poor attribution or failed repair; an average would hide that pattern. These anchors are an initial instrument for calibration and reliability testing, with no validated overall pass threshold.