Manuscript review process

Last updated: 17 October 2026

Goals

• Provide rapid, rigorous review.
• Use expert reviewer time carefully.
• Minimize the number of sequential approval gates where a weak link can stall everything.
• Keep authors, reviewers, and editors clear on who is responsible for the next step.
• Start with 4-month submission-to-publication target, and work toward 2-month target by the end of year 1.
• Treat the first year as an experimental period, with the process revised as we learn.

Summary of the process

1. Authors submit a manuscript.
2. AI scope review and AI-text scan prepared.
3. Managing editors can desk-reject obviously low-quality submissions.
4. Senior editor assigned by managing editors based on expertise.
   • If the assigned senior editor believes another senior editor would be better, they can re-assign themselves.
5. Senior editor can desk-reject obviously low-quality submissions.
6. The senior editor recruits an action editor, whom they supervise, to handle the manuscript or, especially in the first few months, acts as action editor themselves.
7. Action editor performs desk review and makes desk decision.
   • If the action editor is not a senior editor, the supervising senior editor has a 3-day window to overturn the decision or suspend it pending discussion. If a senior editor is acting as action editor, the decision is enacted immediately.
8. Action editor recruits reviewers.
9. Reviewers submit initial reports.
10. Authors, reviewers, and action editor discuss freely in an online forum.
    • Action editor acts as moderator, but there is no mandatory gating or rigid discussion rounds.
11. At any point, the action editor can halt discussion and propose accept, revise+resubmit, or reject.
    • Revise+resubmit verdict: discussion pauses until the authors submit revisions, then discussion resumes as above.
    • Accept or reject verdict: the senior editorial board has 3 days to overturn the decision or suspend it pending discussion; absent either, the verdict is enacted.
12. Upon acceptance, the action editor invites one reviewer to write a review summary.
13. Author approves review summary, or requests revisions until they approve it.
14. Paper enters publication production.

General policies

Confidentiality and identities

The review process is semi-confidential.

Submissions under review are public. Once a manuscript passes desk review, it is listed on a public "Papers Under Review" page, and anyone may self-nominate as a reviewer. Reviewer identities and the review discussion nonetheless remain confidential, as described here:

• The action editor knows the identity of each reviewer, and likewise each reviewer knows the identity of the action editor.
• The authors know the identity of the action editor, in particular whether the action editor is a senior editor.
• The action editor and the reviewers do not know the identity of the authors (all submissions are anonymized; see Author anonymity below)
• Managing editors may see the identities of all participants as needed to manage the process, in particular to check for conflicts of interest.
• The other identities are hidden during review:
  • Author identities are always hidden from the action editor and reviewers.
  • Reviewer identities are always hidden from the authors and other reviewers.
• The review discussion is confidential, except that language from the discussion may be incorporated into the review summary with the approval of both the authors and the reviewer chosen to write the summary.

If the paper is accepted:

• the review summary is published with the paper
• the review discussion remains confidential except insofar as it informs material in the review summary
• reviewers are strongly encouraged to sign the review summary, thereby lifting their anonymity, but may decline to sign (though they should give the editor a reason).

If the paper is rejected or withdrawn:

• reviewer identities remain confidential
• the review discussion remains confidential
• the manuscript is removed from the public "Papers Under Review" page; no public record of rejected or withdrawn papers is kept, but outsiders can infer from previous appearance on "Papers Under Review".

AI use by participants

Separate policy doc: AI usage policy

Reviewer and action editor compensation

Separate policy doc: Reviewer and action editor compensation policy

Public discussion and external forums

For the foreseeable future, the journal does not have the resources to moderate public comment threads, so it will not host them. The journal may link accepted papers to external discussion venues, such as the Alignment Forum, as appropriate.

If a paper under review is publicly discussed elsewhere, the action editor may take public comments and criticism into consideration, but the formal decision remains the responsibility of the journal's review process.

Target timelines

The first-year target is to keep all papers under 4 months from submission to publication, falling to 2 months by the end of year 1.

Default internal targets:

• desk review: within 1 week
• reviewer recruitment: within 1 week after desk review
• initial reviews: within 2–4 weeks after reviewer acceptance
• author discussion after reviews: about 1 week
• revision deadline: normally 1–3 months, if revision is needed
• review of revision: about 1 week
• senior editorial board objection period on a final verdict: 3 days
• review summary finalization: about 1 week after provisional acceptance

Process failures and blockers

No one person should be able to stall a manuscript. If a reviewer, author, or editor holds up progress for too long, the action editor or a managing editor should intervene: send a reminder, reset the deadline, replace the reviewer or action editor, withdraw the manuscript, or decide without the late participant. If the authors are unresponsive, the action editor may also treat the silence as withdrawal.

Appeals, complaints, and resubmissions

Separate policy doc: Appeals, Complaints, Corrections, and Retractions

Submissions by senior editorial board members

When a senior editor is an author on a submission, extra rules apply:

• The managing editor assigns an unconflicted senior editor to supervise per the Conflict of Interest Policy. An author cannot be the supervising or action editor of their own paper.
• Acceptance requires explicit approval from a majority of the senior editorial board (the authors are not counted), rather than the mere absence of objection that finalizes other verdicts.
• The published version publicly discloses the author's senior-editorial-board membership.

These extra rules do not apply for authors who are action editors. (Action editors cannot handle their own submissions of course, but no additional approvals or disclosures apply beyond that recusal.)

Roles and their responsibilities

Managing editor

• Desk rejects obvious spam and low-quality submissions
• Assigns each remaining submission to be supervised by a senior editor based on expertise
• Checks for conflicts of interest and low-quality indicators that involve the author’s identity and passes anonymized information on to senior editor
• Monitors the progress of all submissions toward a decision
• Escalates complaints and appeals to the senior editorial board

Supervising editor

• Must be a senior editor
• Desk rejects obvious spam and low-quality submissions
• Recruits an action editor, or acts as action editor themselves
• Oversees the action editor's conduct of the review
• Raises appeals and contentious issues to senior editorial board

Action editor

The action editor may be appointed by the supervising editor, or the supervising editor may fill the role themselves (a senior editor acting as action editor). Once appointed, the action editor manages the manuscript through review. They are responsible for:

• desk review
• recruiting reviewers
• moderating the review discussion
• keeping the process moving by chasing late participants
• synthesizing reviewer feedback into a decision: accept, revise+resubmit, or reject
• composing decision letters
• recruiting a reviewer to write the review summary or, if none are willing, writing it themselves

Reviewers

Reviewers are asked to:

• declare any conflicts of interest to the action editor (see the Conflict of Interest Policy)
• read the paper PDF (reviewers may consult supplementary material but are not required to)
• write an initial report (with scores per the Scoring system)
• participate in follow-up discussion, in particular, defending, clarifying, or conceding points in their initial report in response to the author’s reply

If recruited by the action editor, reviewers may also draft the review summary.

Authors

The corresponding author speaks for the authors collectively: they upload revisions, accept the review summary, and may withdraw the paper. By default this is the author who submits the manuscript, though the role may be delegated to a co-author or reassigned by the action editor. Throughout this document actions taken by "the authors" should be understood as being taken by the corresponding author.

Detailed review process

Submission

Public preprint

Submitted papers must be publicly posted on an approved preprint server — usually the arXiv, though others may be whitelisted (see the list of approved preprint repositories at Scope and acceptance criteria). This reduces spam and reflects the intuition that if authors are requesting that action editors and reviewers invest time reviewing their paper then they should be willing to post it publicly under their name.

Author anonymity

All manuscript PDFs must be anonymized at submission: author names, affiliations, acknowledgements, and other identifying details removed. Citations to the authors' own work should be made in the third person.

Formatting

Authors may write the manuscript in any reasonable format. It should be converted to PDF for the review process. Conversion to journal publication format happens only after acceptance, so authors do not waste time formatting rejected papers.

Communication

Authors may contact the editors before submission for quick guidance, especially on scope.

Submission process

Authors submit a link to a non-anonymized preprint and an anonymized PDF; the journal extracts the title, abstract, authors, affiliations, and keywords for the authors to confirm or correct. Author-identity information (name, affiliation, etc.) is retained for managing-editor use only, e.g., conflict-of-interest checks; it is not passed on to action editors or reviewers. The submitting author agrees to the journal's licensing requirements and declares any use of AI tools when preparing the manuscript (in compliance with the AI usage policy). They may optionally submit

• One or more links to third-party data repositories from the approved list (see appendix)
• Free-form text comments to the editor, which may address
  • suggestions for reviewers and action editors
  • potential conflicts of interest
  • prior-publication history

AI-assisted intake

Immediately after submission, the journal runs an automated AI scope review that classifies the manuscript as in scope, uncertain, or out of scope, with a short rationale. This is shown to editors to assist assignment and desk review.

Every submitted version of the manuscript is automatically scanned for AI-generated text (currently using Pangram). These scans are visible only to editors.

Assignment to senior editor and action editor

Each submitted paper is assigned by a managing editor to a senior editor with relevant expertise. The supervising editor chooses to either:

• act as action editor themselves; or
• appoint an action editor.

In the first few months, senior editors are encouraged to handle papers themselves when feasible. This helps establish shared standards. As the journal grows, senior editors should increasingly delegate to a broader pool of action editors.

The managing editors will assist in identifying possible action editors.

The action editor begins with desk review.

Desk review

Desk review checks whether a manuscript is a plausible enough publication candidate to justify spending scarce reviewer time. Desk review considers:

• fit with scope and other acceptance criteria
• seriousness and completeness
• obvious fatal flaws
• conflicts of interest
• whether there is a suitable senior editor to supervise
• whether suitable reviewers are likely to exist
• whether the claims are interesting enough to review

The action editor records an explicit desk decision — approve or reject:

• If the action editor is not a senior editor, their proposed desk decision will be sent to the supervising editor, who will have 3 days to overturn it, or to suspend it pending discussion. If they approve the decision or do nothing, the decision stands.
• If a senior editor is acting as action editor (i.e., self-supervised), the decision is immediately enacted.

A manuscript advances to reviewer recruitment only on desk approval, which is also what lists it on the public "Papers Under Review" page and opens it to reviewer self-nomination.

To reduce the burden on editors, they may consult the automated AI scope review. It is advisory and should be considered unreliable; the editor should not reject the paper unless they can confirm the purported issue themselves.

Reviewer recruitment

The action editor recruits reviewers with suitable expertise and interest.

The action editor should normally target 2–3 reviewers for each manuscript. Four reviewers may be appropriate for unusually broad, interdisciplinary, or disputed papers. A decision based on a single reviewer will only be made in rare circumstances where the action editor has engaged closely with the manuscript themselves, and has the agreement of the supervising senior editor.

Candidate reviewers may be found through traditional sources: editor professional networks, the journal’s reviewer database, internet search, and directly from the literature. They may also be sourced from non-standard sources that we are developing:

• Our review webapp hosts an AI assistant that is tailored for reviewer discovery; it is integrated with Semantic Scholar, has powerful skills for finding email addresses on the web, and allows two-click invitations.
• The journal website hosts a public "Papers Under Review" webpage that will be regularly promoted on social media, and any researcher may self-nominate there.
• In the future, we may integrate with the Alignment Forum to "cross-post" the under-review list and allow self-nomination of users there (but this has not been developed yet).

Review delegation

Reviewers should not delegate review duties to a student or other collaborator while posting under their own name. Instead, they should give the action editor the student's email address so that, if the action editor agrees, the student can be invited to participate in the discussion as an additional reviewer.

Conflicts of interest

Reviewers should be asked to disclose conflicts of interest before accepting. Because authors are anonymized, a reviewer often cannot identify an author-based conflict. The managing editors — who can see author identities — are responsible for screening reviewer–author conflicts. Reviewers should still disclose any conflict they can detect from the manuscript or otherwise know.

Initial reports

Each reviewer submits an initial report before seeing any other reviewer's report, the discussion between authors and reviewers, or any AI-generated review supplied by the journal. Submitting their own report is what gives a reviewer access to the discussion.

A report should normally include:

• short summary of the paper
• main strengths
• main weaknesses
• questions for authors
• requested changes, if any
• recommendation: accept, revise+resubmit, or reject
• numerical scores (see Scoring system)
• AI-use disclosure

Reviewers should distinguish cleanly which issues

• must be fixed to permit publication
• are optional but, if not addressed, would be highlighted as deficiencies in the review summary
• are strictly helpful suggestions

They should not treat the review as a list of everything that could improve the paper, but focus on what is needed for publication according to the acceptance criteria (Scope and acceptance criteria).

Review discussion

After initial reports are submitted, authors and reviewers may discuss the paper: correcting misunderstandings, settling factual disputes, and establishing which changes are necessary for publication. The action editor moderates the discussion and may close threads that are no longer useful.

Unlike a traditional journal, the discussion is not divided into rounds gated by the editor: it runs continuously, and the action editor ends it when it stops being productive or when they have enough information to decide.

Manuscript decision process

Possible outcomes. After the action editor has collected enough information from the review discussion, they make one of three decisions:

• Revise+resubmit: If the paper is promising but not yet publishable, the action editor may ask the authors to revise the manuscript to satisfy specific demands.
  • The action editor should distinguish required changes from recommended but optional improvements.
  • Revision deadlines are negotiated with authors. The default should usually be 1–3 months, depending on the requested changes. If the author misses the deadline, the editor may withdraw the manuscript.
  • Following resubmission, the action editor may choose to re-open discussion with the reviewers or instead may make a final accept/reject decision themselves.
  • In rare cases, this may repeat by the action editor requesting additional revisions, but multiple rounds should be avoided when possible. The action editor should avoid expanding the required-change list during later rounds unless new information reveals a serious issue.
• Acceptance (see below). The review summary process begins and the paper proceeds to publication.
• Rejection (see below). The paper is removed from the public "Papers Under Review" list and the review process is stopped.

All accept/reject verdicts go to the senior editorial board. Once a submission has advanced beyond desk review, every accept/reject verdict is raised to the entire senior editorial board before it is finalized. The action editor proposes a verdict, and the senior editors have 3 days to object. Non-response counts as no objection. If no one objects, the verdict is finalized; if anyone objects, it goes to senior-editor discussion and, if needed, a simple-majority vote of the non-conflicted senior editors. This is a deliberately conservative policy for the journal's early days, and the journal expects to relax it as it develops.

Authors' options after a decision — appeals, complaints, and resubmission of a rejected paper — are described in Appeals, Complaints, Corrections, and Retractions.

Withdrawal by the authors

The authors may withdraw the manuscript at any time: the paper is removed from the public "Papers Under Review" list, no further revisions or decisions are recorded, and the review participants are notified.

Withdrawal of an already-published paper is handled as a retraction; see Appeals, Complaints, Corrections, and Retractions.

Decision letters

Decision letters should be direct and useful. They should state the decision and the main reasons. If it is a revise+resubmit decision, they should state any required changes, whether another round of review is expected, and the relevant deadlines.

Review summary

Each accepted paper is published with a review summary (Review summary). The action editor invites one reviewer to write it. If the invited reviewer declines or submits a summary that is not usable, the editor may invite another. In rare cases where no reviewers are suitable and agree, the action editor should write the summary themselves.

If authors have objections to the draft review summary, they may request changes. (This essentially reverses the reviewer and author roles relative to the main paper.) If acceptable wording cannot be reached, the authors may withdraw their manuscript before publication. This is very costly, and the action editor should work hard to find wording that both reviewers and authors can live with.

All reviewers (not only the one who wrote it) are encouraged to publicly sign the review summary once it is finalized. However, they may decline to sign, sign anonymously, or, at the editor’s discretion, sign with a pseudonym.

Post-acceptance production

After official acceptance:

• authors provide source files (e.g., LaTeX) or other materials needed for publication
• the journal produces the published versions and indexing metadata (including DOI)
• the review summary is finalized
• reviewer identity and signing choices are finalized
• reviewer/action editor's compensation is processed under the Reviewer and action editor compensation policy

Published papers and review summaries are released under the journal's publication license.

Appendix

Approved preprint and data repositories

Managing editors should consult with the senior editors when adding additional repositories to the approved list, but for expediency they may immediately add something to the list and alert the senior editors after the fact for approval or reversal.

Approved data repositories

Supplementary material may be submitted as links to third-party data repositories on this approved list:

• Zenodo (zenodo.org)
• GitHub, GitLab, and Anonymous GitHub (github.com, gitlab.com, anonymous.4open.science)
• FigShare (figshare.com)
• Dryad (datadryad.org)
• OSF (osf.io)
• Code Ocean (codeocean.com)
• Hugging Face (huggingface.co)
• Harvard Dataverse (dataverse.harvard.edu)
• Google Colab (colab.research.google.com)