On this page

Reviewer participation, AI assistance and transparent review history
Prospective notice dated 29 October 2026 (Asia/Singapore). Version: jol-reviewer-notice-2026-10-29-v1.

Fully agentic editorial workflow with human peer reviewers (at least 2) and Council of LLMs as peer reviewers.

The Journal of Learnomics is published by IAIED, Singapore UEN 53487769M, 60 Paya Lebar Road, #07-54 Paya Lebar Square, Singapore 409051. Contact editor@mylearnomics.com. AI editorial profiles describe software roles, not human scholars; the deployed pilot combines workflow software, a three-model council and documented decision rules. Not every listed persona is a separately deployed agent. The publisher remains accountable for verifying reviewers, conflicts, exceptions, appeals and publication.

Choosing to receive invitations

Joining the reviewer roster is voluntary and separate from subscribing to author calls or newsletters. An account holder may explicitly opt in through the reviewer portal to manuscript-specific invitations relevant to verified expertise. Public authorship records or a place in a discovery catalogue do not establish consent, verified identity, suitability or appointment. The publisher verifies identity and expertise, checks conflicts and assigns reviewers only after the current manuscript passes documented editorial clearance.

You may decline any invitation. You may withdraw future invitations in the portal or through an invitation’s withdrawal link; simply opening the confirmation page does not change your preference. Withdrawing invitations does not silently cancel an already accepted assignment: contact the editorial office if you cannot finish an active review. Necessary messages about an accepted assignment remain separate from future invitation preferences.

Independent, double-anonymous assessment

Acceptance requires at least two independent human reports on the current manuscript round. AI reports never count toward this minimum. The publisher knows reviewer identities for verification and accountability; the review copy and reports released to authors are anonymised. Public preprints, citations or specialist knowledge can make authors recognisable, so double-anonymous procedures cannot guarantee unidentifiability. Declare any actual or potential conflict, including recognition of an author, and do not attempt to uncover identities.

Accept an assignment only if you have the expertise and time to evaluate it fairly. Assess the research question, methods, evidence, reporting, limitations and conclusions yourself. Keep review materials confidential. Do not upload unpublished manuscripts, reviews or private correspondence to public or unapproved AI services. Put sensitive integrity concerns in the separate confidential-to-editor field, not in text intended for authors or publication.

Optional Council-assisted drafting

Where the reviewer portal offers assistance, first read the current manuscript and record your independent initial assessment before viewing an AI-assisted draft. Assistance is optional. It requires the author’s separately recorded permission for anonymised manuscript AI processing and the reviewer’s explicit opt-in for the specific assistance. The reviewer can edit or discard the suggested text; it is not automatically submitted as a human report. Verify every substantive claim, reference and recommendation you adopt, disclose the assistance used, and personally approve the final assessment.

The council uses three Google Gemini variants from the same provider and model family. Their errors may be correlated; they are not three independent human experts. Google Cloud hosts the journal service and result database in Singapore, but the model endpoint is global. Consented model requests use the supplied anonymised title, abstract, keywords and manuscript plain text, and only reviewer text explicitly approved for that purpose. Uploaded PDFs, private author fields and confidential editor comments are excluded. Remove identifying and unnecessary sensitive details from supplied text itself. Initial assessment, generated suggestion and final-report versions are recorded for accountability; AI processing and later public publication are separate choices.

Revision rounds

Authors respond to numbered reviewer and editorial points, explaining each change with its manuscript location or giving a reasoned response where they disagree. A revised submission is a new recorded round and must pass current clearance. The clean anonymised revision, response and any supported change record are assessed against the concerns raised. Prior-round reports do not replace the two required current-round human reports. The publisher may seek additional independent expertise or hold the process where concerns remain.

Optional publication of review history

For an accepted article, the journal can prepare a separate review-history package containing dated review rounds, anonymised human comments to authors with separately approved final rubric scores, rationales, evidence anchors and validity/integrity flags, clearly labelled AI council reports and model information, author responses, and public editorial decision summaries. It remains private until the publisher completes confidentiality and rights checks and the author approves the exact package for publication. Review reports are distinct from the article; agreement to review, receive invitations or use AI does not itself grant permission to publish a report.

A human report is included only with that reviewer’s separate affirmative publication permission for the exact final public report and final submitted rubric shown for approval. Attribution uses an anonymous reviewer label and manuscript/round reference; reviewer names, email addresses, affiliations, account identifiers and signatures are excluded. No confidential-to-editor comments, raw audit logs, private API payloads, participant-identifying information or unauthorised third-party material may be included. Initial notes and AI-assisted drafts are not automatically part of the public package. Only final reports, final submitted rubric scores with their rationales, evidence anchors and validity/integrity flags, recommendations and AI-assistance disclosures explicitly shown and approved for publication are eligible. The reviewer’s initial notes and initial rubric remain private. Public rubric summaries are descriptive human judgments, not validated measures or acceptance thresholds; AI model scores are shown as not scored unless a separately identified, verified model-scoring feature actually supplied them.

The publisher retains the private original and prepares any necessary redacted public copy with a recorded reason. Redactions must protect confidentiality or rights without concealing substantive criticism. Any edited public copy requires the reviewer’s approval of that exact combined report-and-rubric version. Final rubric content is not silently rewritten: any necessary correction must follow a documented process, and identifying or otherwise unsafe text blocks release until resolved. If permission or approval is absent, the report remains private. The package must state how many human reports were included and withheld and describe itself as selected peer-review history when incomplete; it must not suggest that an omitted report was favourable or that all reviews are public.

Reviewers retain their rights in their report. If you opt in, the approved public report and final rubric are offered under Creative Commons Attribution 4.0 International (CC BY 4.0), using the anonymous attribution specified above. That licence permits redistribution and adaptation, including commercial reuse, subject to its terms. You may withdraw unpublished permission by contacting the journal before release. Once published under CC BY, compliant users’ granted rights cannot be revoked and independent copies cannot be guaranteed to disappear. The journal addresses later errors or integrity concerns through a documented correction, withdrawal or other appropriate notice rather than silently rewriting the record. AI-generated content is identified as such; any stated licence applies only to rights the journal is entitled to grant.

Declining optional public publication does not disqualify an otherwise valid human review and does not change the scientific acceptance standard. Earlier reports are not retrospectively licensed. Authors separately approve the exact public package and the publication rights for their response text; their approval cannot authorise another person’s private report.

Data, email and retention

The journal records account/contact details, verified expertise and identity references, conflicts, assignments, invitation preferences, reviews, assistance disclosures, permissions and audit records. WordPress and the editorial mailbox use HostGator. Permitted invitation and workflow emails use Twilio SendGrid, processing recipient addresses, headers, subjects and message content through the configured United States route. Invitations can contain the anonymised manuscript title, abstract, due date and portal link; manuscript PDFs and full proofs are not attached. Only separately opted-in manuscript-specific reviewer invitations qualify for the reviewer invitation route. Unconsented author campaigns, unknown categories and unclassified journal messages remain blocked from that transport.

The adopted retention schedule, jol-retention-2026-10-28-v1, provides annual roster review and removal of inactive profiles after two years without engagement unless participation is renewed, retaining justified opt-out suppression information. Necessary evidence in an article’s review file follows that file’s schedule: seven years after publication or final appeal closure, whichever is later, for accepted papers; two years after the final decision or appeal closure for rejected or withdrawn submissions. Public scholarly records and necessary rights evidence may remain longer. Operational logs/backups have the adopted 90-day schedule with documented exceptions; this is publisher-administered and does not claim all copies are automatically purged. Cloud job results expire after 30 days, with asynchronous Firestore deletion; WordPress copies and provider retention are separate.

Send access, correction, withdrawal, confidentiality or retention requests to editor@mylearnomics.com. The publisher may verify identity and explain applicable limits, including completed processing, the public scholarly record and documented holds. This prospective notice does not establish anyone’s assent by itself. Features and choices are available only when shown as enabled in the portal. See the privacy notice, editorial workflow and publication terms.