AIC Preprints

About the Preprint Server

A preprint is a complete piece of research made public before, or alongside, peer review. The practice has been standard in physics, mathematics and computer science since arXiv began in 1991, and has since spread to biology, medicine, chemistry and the social sciences. It exists because research sitting in a review queue for eight months helps nobody, and because the author who circulated an idea first should be the one credited with it.

AIC Preprints applies that model to the Caribbean. It is a regional server in the tradition of arXiv, and closer still in spirit to AfricArXiv, which was built on the argument that a region should own the infrastructure that holds its scholarship rather than rent it from elsewhere. Caribbean researchers are often working against a hurricane season, a budget cycle, a health emergency or an election, any of which can close the window in which a result is worth acting on.

Reasons to Post a Preprint

• Priority. The announcement date is a permanent public record that the work was yours on that day.
• Speed. Two working days from submission to a citable identifier, against several months for a reviewed article.
• Feedback before commitment. Readers find errors that reviewers would have found later, while there is still time to fix them cheaply.
• Funder and thesis compliance. Many grants and postgraduate programmes now require open deposit of outputs.
• Reach. Preprints are indexed, discoverable and free to read from any connection, including the ones that cannot afford a subscription.
• No penalty at this journal. Posting a preprint here or anywhere else does not disqualify a manuscript from AIC peer review. It never has.

Scope and Eligibility

The server accepts research that applies machine learning or artificial intelligence to the Caribbean, to Small Island Developing States, or to small-state conditions the region shares. Every preprint carries one primary code and up to three cross-list codes from the AIC subject classification, which runs on two axes: fifteen Caribbean subject areas and twelve method tags.

We accept research articles, data and methods notes, replication reports, working papers, theses and dissertations, conference papers, and technical reports. We accept manuscripts already under review elsewhere and manuscripts already published, provided the author holds the right to deposit the version submitted.

Following the practice arXiv adopted for computer science in 2025, we do not accept review articles or position papers on the preprint server unless they have already passed peer review at a journal or conference. Unreviewed surveys are the largest source of low-value submissions on every server that permits them. Send an unreviewed review article to the journal instead, where it goes to reviewers.

Registration Is Required

Until January 2026 most preprint servers treated an institutional email address as sufficient evidence that a submitter was a working researcher. arXiv stopped, and the servers that did not follow are now absorbing machine-generated submissions at a rate that buries the real work. An unauthenticated server drowns. That is the entire argument for making you register, and it is not a theoretical one.

A registered account establishes that a named, contactable person stands behind the submission, carries your endorsement status so you are not re-endorsed every time, holds your declarations so they are confirmed rather than retyped, and gives us somewhere to attach a rate limit that is generous for a human and useless to a script. Reading the server requires no account and never will.

The arXiv-Style Checks

Every submission passes through two rounds of checks, and it is worth knowing which is which. The account checks are applied once, when you register. The moderation checks are applied to every manuscript, every time.

Neither round assesses whether your findings are correct. That is what peer review is for, and it is why a preprint record says on its face that no reviewer has vouched for it.

Run These Checks Yourself First

Nearly every rejection on this server is one of the following, and every one of them is cheaper to catch before you upload than after.

1. Is it a review article or a position paper that has not been peer reviewed? If so it does not belong here. Send it to the journal instead, where it goes to reviewers.
2. Does the manuscript say what is Caribbean, small-state or SIDS about it? A reader should not have to infer it from the author’s address.
3. Are all your references real, and have you read them? Fabricated citations are the single most common integrity failure on every server, and a moderator checks them in minutes.
4. Is there an AI use statement? Tool, version, task. “None” is a perfectly good answer when it is true.
5. Are funding, competing interests, ethical approval and data availability all stated? All four, not three.
6. Can anyone be identified in your data, figures or quotations? Apply the small-population test in the research ethics policy rather than a threshold designed for a country of eighty million.
7. Do you actually hold the rights? Check your publisher agreement, and check the licence on every figure you did not draw yourself.
8. Are author names visible? Preprints are not anonymised. A file with the names stripped out is returned.
9. Is the licence the one you meant? It is irrevocable once the version is announced.
10. Have you chosen the right primary subject code? Guessing is fine and is corrected without penalty; leaving it wrong on purpose is not.

Endorsement

First-time submitters need an endorsement. This is the mechanism arXiv uses to establish that a submitter is working within a scholarly community, and it is why arXiv stopped treating an institutional email address as sufficient on its own in January 2026. Automated and machine-generated submissions almost all arrive from first-time posters, and an unendorsed server drowns.

You do not need an endorsement if any of the following applies:

• You have previously had a preprint announced on this server
• You have published a peer-reviewed article in AIC or in any indexed journal
• You are a member in good standing of the Caribbean Artificial Intelligence Association
• A co-author on the submission meets any of the above

If none applies, name a proposed endorser on the submission form, or ask the moderation panel to find one for you at preprints@caribbeanaijournal.org. We will not leave a genuine researcher stranded for want of a contact, and endorsement is free.

Students: secondary school, sixth form and undergraduate authors need no endorsement. Use the Student Researcher Section, which returns written reviewer feedback that the preprint server does not provide.

From Submission to Announcement

Moderation Criteria

Moderators assess only whether a submission belongs on the server. They do not judge whether the findings are right, and they do not provide reviewer-style feedback. A submission is accepted when it meets all seven of the following.

Rejected submissions are not announced. The decision names the criterion that failed, without a critique of the work. You may appeal once, in writing, to the Editor-in-Chief. Appeal decisions are final. Repeated rejections lead to closer scrutiny of later submissions from the same author.

Machine-Generated Manuscripts

Every preprint server in operation is now handling submissions written by language models and presented as original research. Use AI tools as freely as you use a compiler or a statistics package, and describe that use in the manuscript: which tool, which version, for which task. A manuscript whose argument, analysis or results were produced by a model that no author has verified is rejected. If a moderator cannot establish that a human stands behind the claims, the submission does not go up. The generative AI policy has the full text.

Licensing

You keep copyright in everything you post here. When you submit, you choose one of four licences, and your choice is displayed on the preprint record.

All four are irrevocable for versions already announced, which is what makes the record permanent enough to cite.

Withdrawal

Announced preprints are not deleted. If work is superseded, wrong, or withdrawn for any other reason, submit a withdrawal notice and we post it as a final version explaining the withdrawal. The identifier keeps resolving, the earlier versions stay readable, and anyone who cited the work can see what happened. Records are removed outright only where the law requires it, for example a confirmed copyright infringement or an unlawful disclosure of personal data.

Document Control

Preprints and the Journal

The two tracks are separate. Posting a preprint carries no obligation to submit it to the journal, and submitting to the journal carries no obligation to post a preprint. You may do both on the same day. If a preprint is later accepted for an edition, the peer-reviewed version of record is published in that edition and cross-linked to the preprint, and the preprint record notes that review took place.

Moderation of a preprint gives it no standing in peer review. Editors and reviewers assess a submitted manuscript on its merits and are not told that a preprint version exists unless the authors mention it themselves.