Moderation & integrity

The moderation queue

Studies and datasets share one queue, one set of actions, and one state machine. A moderator can assign a reviewer, approve, or reject, and every one of those actions moves the publication through the same validated transitions that the collaborative review system uses. There is no side door that sets a status directly.

The audit trail

Every publication transition writes two things: a transition record on the item, and an append-only audit entry naming the actor, the action, and the metadata. The same applies to review decisions, comment sharing and unsharing, reviewer assignment, and licensing events.

• Authors can see the transition history of their own work.
• Reviewers can export the review history of an item as JSON or CSV.
• Audit entries are not editable from the application, by design.

Raising a concern

1. 1

   Comment first, if it is a scientific question

   Disagreement about method, statistics, or interpretation belongs in public discussion on the study. That is what the record is for.

2. 2

   Report if it is an integrity matter

   Plagiarism, undisclosed conflicts, fabricated or duplicated results, missing ethics approval, or data published without consent should be reported rather than argued in comments. Use the report control on the item, or write to [email protected].

3. 3

   Give evidence, not a verdict

   Identifiers, the specific claim, and what you compared it against. Reports without anything checkable are slow to act on.

4. 4

   Expect a human

   Integrity reports are handled manually. Authors are given the chance to respond before anything is changed, unless there is an immediate legal or privacy risk.

Corrections and retraction

Publishing has exactly one exit, and it is a loud one. A published study can never revert to draft or disappear once other people have started citing it; the only move out is aretracted state that requires a reason and a public notice. Everything else is done in the open, as a new version or a correction.

New version

The normal route for updated analysis, added subjects, or fixed figures. The old version stays addressable.

Correction notice

For an error that changes interpretation but not the conclusion. Attached to the study so readers see it.

Retraction

For work that should no longer be relied on. The page and its DOI keep resolving to the retraction notice, because removing it silently would break every citation pointing at it — but the study leaves every listing, search result and leaderboard, stops counting toward its authors’ verified publications, and can no longer be downloaded.

Takedown

Reserved for legal or privacy obligations — copyright claims, or personal data published without a lawful basis.

How a retraction happens

• The corresponding author retracts their own study from the study page, or from My Submissions in the desktop app. They choose a reason — author request, honest error, a problem with the data, an ethics or consent problem, duplicate publication, a legal obligation — and write the notice readers will see.
• Misconduct is the one reason an author cannot select. It is a finding about people, not a statement about the work, so only the BCINexus editors record it.
• Every retraction is notified to the co-authors and the editors, appended to the study’s transition history, and written to the audit log. The original decisions and reviews are never rewritten.
• Corrected work is not stranded: a retracted study can go back to draft, be fixed, and be submitted for review again. The retraction stays on its record.

Withdrawing before publication

A submission that has not been published yet is still yours to pull back. Withdrawing stops the review, tells the reviewers who were on it, releases their assignments, and returns the study to draft — nothing you have already received is deleted, and you can submit it again when you are ready.

What is not allowed

• Human-subject data published without consent or an ethics basis.
• Work you do not have the right to publish, including data owned by a former institution.
• Fabricated results, or metrics reported from a split that leaked.
• Attempting to review your own submission, or arranging reciprocal approvals.
• Using material seen under review before it is published.

The full terms are in the acceptable use policy, community guidelines, and the reviewer agreement.

Related