Publication Ethics

            KJAI is committed to maintaining the highest standards of publication ethics and academic integrity, and follows the ethical publishing policies of The Korean Academy in alignment with internationally recognised standards and the guidance of the Committee on Publication Ethics (COPE).
          

The journal operates a rigorous double-blind peer review process to ensure fairness, objectivity and confidentiality. All submissions are evaluated solely on scholarly merit, originality and relevance to the journal's scope — never on the author's institution, nationality, career stage, employer, or connection to the publisher.

Ethical standards cover, but are not limited to:

Originality and the avoidance of plagiarism
Proper authorship and acknowledgement of contributions
Disclosure of conflicts of interest, including commercial and funding relationships
Ethical conduct of research involving human participants and human-derived data
Lawful and licensed use of data, models and code
Data transparency, code availability and research integrity
Disclosure of generative AI use in research, analysis or writing
Honest reporting of results, including negative and inconclusive ones
Consideration of foreseeable misuse and differential harm
Confidentiality in peer review
Responsible and independent editorial decision-making
Handling of complaints, corrections and misconduct

All participants in the publication process — authors, reviewers, editors and the publisher — are expected to comply with these standards.

Ethics Oversight

Ethical matters are referred in the first instance to the Editor-in-Chief, who is responsible for the integrity of what the journal publishes. Until that appointment is made, this responsibility is held by the Executive Editor together with the publisher's Academic Publishing Division. Where the responsible editor has a conflict of interest, or where the matter concerns them, it passes to the publisher, The Korean Academy, which appoints an independent person with no involvement in the case to consider it.

Where an author, reviewer or editor is affiliated with a university, company or other institution, serious allegations may also be referred to that institution's research integrity office. Every ethics decision, and the reasons for it, is recorded in writing.

Research involving people and their data

Much of what this journal publishes is trained on, evaluated on, or deployed upon data that came from people. Where research involves human participants, human-derived data, or the deployment of a system that affects people, authors must state the consent basis and, where their institution required it, the ethics approval and its reference number. Where a public dataset contains personal data, the licence and the terms under which it was collected still apply — a dataset being downloadable is not the same as its subjects having agreed to your use of it.

No participant may be identifiable in a published article, including through a figure, a quoted example, or a model output reproduced in the text.

Dual use and foreseeable harm

Applied AI research can be repurposed. Authors are asked to consider, and where relevant to state, the foreseeable misuse of what they publish and the populations on whom errors fall unevenly. The journal will not publish work whose primary purpose is surveillance of a population, deception, evasion of safety controls, or harm to a person or group, and reserves the right to ask for a specific detail to be withheld where publishing it would create a concrete risk that outweighs the scientific value of including it. That is a conversation with the authors, not a unilateral edit.

A note on this journal's particular position

          The Korean Academy is a teaching institution, and some KJAI authors and reviewers will be its students, alumni or staff. Rather than pretend otherwise, the journal declares these connections and removes the conflicted person from the decision.
          
          Students of The Korean Academy may submit their manuscripts to this journal, and are welcome to. What follows is identical to what follows for anyone else: double-blind review by two independent reviewers, with the manuscript assigned to an independent handling editor who selects those reviewers and takes the decision alone. Acceptance is determined by the outcome of that review and by nothing else. Submitting as a KORA student confers no advantage, and the connection to the publisher confers no disadvantage.
          
          Equally, no KORA student is required or expected to publish here. Any coursework requiring a research project may be satisfied by submitting to any appropriate venue, or to none.

Ethics Statement

The duties below apply to everyone involved in publishing an article in KJAI. They are drawn from COPE's core practices and are sent to every author on acknowledgement and to every reviewer on invitation.

Duties of

Authors

Originality. Submit only original work that is not published elsewhere and is not under consideration by another journal. A posted preprint is declared, not concealed. Text reused from your own earlier work must be quoted or paraphrased and cited like any other source.
Authorship. Everyone listed must have contributed substantively and approved the submission; everyone who qualifies must be listed. Contributions are declared using CRediT terms on the title page. No AI system is named as an author.
Data and result integrity. Present genuine data and genuine numbers. Fabrication, falsification, selective reporting of runs or seeds, and reporting a tuned result as a held-out one are misconduct. Retain your raw results and be prepared to supply them if asked.
Reproducibility. Provide a code and data availability statement, complete the reproducibility checklist, and report the setup — splits, hyperparameters, hardware, seeds, compute — well enough that the work could be repeated.
Lawful and licensed use. Use data, models and code within the terms of their licences, and say which licences those are.
Research ethics. Obtain informed consent and, where your institution requires it, ethics approval — before collecting data. No participant may be identifiable in the published article.
Disclosure. Declare funding, conflicts of interest including commercial ones, any connection to The Korean Academy, and any meaningful use of generative AI. Language polishing counts as meaningful use.
Permissions. Obtain written permission for third-party material — figures, images, long quotations, instruments — before submission.
Honest claims. Do not state in the abstract what the results do not demonstrate. Report limitations and foreseeable harms rather than waiting for a reviewer to find them.
Correction. If you discover a significant error in your own published work, tell the journal promptly and cooperate in correcting the record.

Duties of

Editors

Independence. Decide on scholarly merit alone. The publisher supports production and dissemination; it does not influence editorial decisions or the peer review process.
Fairness. Judge manuscripts without regard to the author's race, gender, sexuality, religion, nationality, institution, employer, career stage or political outlook.
Matching expertise. Send cross-domain manuscripts to reviewers who between them cover both the method and the domain, and obtain a third opinion where neither can assess a critical part.
Confidentiality. Disclose nothing about a submitted manuscript, or the code and data supplied with it, to anyone other than the author, the reviewers and the editorial office.
Conflicts. Step aside entirely from any manuscript you are connected to — no reviewer selection, no access to reports, no decision. This applies most strictly to an editor's own submissions.
Separation of fees and decisions. Never raise an invoice before a decision to accept, and never learn whether a fee has been paid or waived while handling a manuscript.
Accountability. Investigate credible allegations whenever they arise, including after publication, and correct the record where needed.

Duties of

Reviewers

Confidentiality. Treat the manuscript, and any code or data supplied with it, as unpublished and confidential. Do not share it, discuss it outside the editorial process, or use its content before publication.
No AI upload. Do not upload a manuscript, its data or its code — in whole or in part — to any generative AI service or hosted model. This is a breach of confidentiality, not a grey area.
Competence. Accept only what you are qualified to assess, and tell the editor which parts of a manuscript fall outside your expertise. Partial competence is expected in a cross-domain journal; silence about it is not.
Conflicts. Decline where you have collaborated with, supervised or been supervised by a likely author within three years, share a department or company, have a close personal relationship, or stand to gain from the outcome. Declare commercial interests.
Objectivity. Support every criticism with an argument and a location in the text. Personal criticism of an author is not acceptable, and neither is rejecting a sound study for reporting a modest or negative result.
Promptness. Return the report within 21 days, or decline early so the editor can approach someone else.
Alertness. Report suspected plagiarism, duplicate publication, fabricated data, missing consent, undisclosed conflicts or undisclosed AI use to the editor — describing what you saw, without accusing or investigating.

Duties of

The Publisher

Non-interference. The Korean Academy supports the journal's production, dissemination and preservation, and does not influence editorial decisions or peer review.
Support for integrity. Provide the similarity-checking, record-keeping and editorial resources that ethical publishing requires.
Transparency. Publish fees, licensing, ownership and editorial policy openly, and never claim an index, identifier or metric the journal does not hold.
Preservation. Maintain the published record so it survives changes to journal management or hosting.
Escalation. Appoint an independent person to consider any matter in which the responsible editor is conflicted.

Generative AI Policy

This journal publishes research about artificial intelligence, so it would be absurd to forbid its use. The policy is therefore about disclosure and responsibility, not prohibition: you may use these tools, you must say that you did, and you remain answerable for every word and number in your paper.

          The three rules
          1. Declare it. Name the tool, the version, and what it was used for, in the Declarations block. Language polishing counts.
          
          2. An AI system cannot be an author, cannot be credited as one, and cannot hold responsibility. Authorship requires accountability, and a model cannot be accountable.
          
          3. You are responsible for everything. If a tool produced a citation, a number, a translation or a sentence that turns out to be wrong, that is your error, handled as any other error would be.

For authors

Writing assistance — drafting, editing, translation, summarising, code comments. Permitted, and must be declared.
Generative AI as part of the method — a model used to label, generate, augment, evaluate or decide. Permitted, and must be described in the Method section with version, settings, prompts or a representative sample, and access date. Hosted models change without notice; a result is only interpretable if the reader knows which model, and when.
Fabricated citations or results. Never acceptable. Check every reference a tool suggests against the actual source; a citation to a paper that does not exist is treated as a serious integrity failure.
Unexamined black boxes. A manuscript that runs a generative model over a dataset and reports the output with no evaluation of its quality is returned at screening. Using a model is not a finding.
Images and figures. Generated or AI-modified images must be identified as such in the caption. Never use a generative tool to alter a figure that reports data.
Confidential data. Do not paste participant data, clinical records or anything else you hold under a confidentiality obligation into a hosted model.

For reviewers

          Do not upload a manuscript, its data or its code to any generative AI service or hosted model. A manuscript under review is confidential and is not yours to disclose; sending it to a third-party service discloses it. This includes pasting an abstract into a chatbot to help you summarise it.
          
          You may use an AI tool to polish the language of a report you have written yourself. Declare it to the editor. The judgements must be yours: a report that is visibly generated rather than written is not a review, and is not used.

For editors

Editors do not use generative AI to make or draft editorial decisions, and do not submit manuscripts or reports to hosted services. Administrative use — scheduling, formatting, correspondence templates — is permitted where nothing confidential is disclosed.

What happens if AI use is not declared

Undisclosed use is handled under the Malpractice Policy. Before publication, that normally means the manuscript is returned for a corrected declaration; where the undisclosed use is substantial and misrepresents the work, it means rejection. After publication it means a correction, or a retraction where the finding itself cannot be relied on. As with any allegation, the authors are shown the evidence and given at least 21 days to respond before a conclusion is reached.

Malpractice Policy

This policy sets out what KJAI treats as research and publication misconduct, how an allegation is handled, and what follows if it is upheld. It applies to submitted and published work alike, and there is no time limit on investigating a credible allegation.

What counts as misconduct

Plagiarism — using another person's words, ideas, data, figures or code without attribution, including material translated from another language.
Self-plagiarism — reusing substantial parts of your own published work without quotation and citation.
Duplicate or redundant publication — publishing the same work more than once, or submitting to more than one journal at a time. A declared preprint is neither.
Fabrication and falsification — inventing data or results, altering numbers, or manipulating images.
Selective reporting — omitting findings, runs or seeds that do not support the argument, or reporting the best of many runs as though it were typical.
Misreporting evaluation — reporting a tuned result as held-out, using the test set to make design decisions without saying so, or comparing against baselines given a materially smaller search budget than your own method.
Licence and data violations — using data, models or code outside the terms of their licence, or using personal data without a lawful basis.
Citation manipulation — citations added to inflate a metric rather than to support an argument.
Authorship misconduct — gift, guest or ghost authorship, or omitting someone entitled to authorship.
Undisclosed conflicts — including undeclared funding, an undeclared commercial interest in a system being evaluated, or an undeclared connection to The Korean Academy.
Undisclosed generative AI use, fabricated citations produced by a tool, or listing an AI system as an author.
Ethics breaches — research with human participants or human-derived data conducted without required consent or approval.
Peer review misconduct — fake, coerced or manipulated review; breaching confidentiality; uploading a manuscript to a generative AI service; misappropriating content, data or code from a manuscript under review.
Paper mills — manuscripts produced or brokered for sale.

Similarity screening

Every submission passes a similarity check before it goes to review. A similarity score is a starting point, not a verdict: a high score caused by quoted material, standard method descriptions or a well-referenced literature review is not plagiarism, while a low score containing one stolen paragraph is. Reports are read by an editor, who looks at what overlaps and whether it is attributed.

The journal does not use an "AI detector" to accuse anyone of anything. These tools are unreliable, and they misfire most often on authors writing in a second language. Where AI use is suspected, the question put to the author is a factual one about what they used, not an accusation derived from a score.

How an allegation is handled

Recording. The concern is recorded in writing and assessed by an editor with no connection to any party. Allegations are considered on the evidence, whoever raises them — including anonymously.
Notification. The corresponding author is told what has been alleged, shown the evidence, and given at least 21 days to respond. No conclusion is reached before that response is read.
Assessment. The response is weighed against the evidence. Further expert opinion may be sought, and the manuscript may be put on hold meanwhile.
Institutional referral. Where the response is inadequate and the matter is serious, we contact the authors' institution or employer. The journal investigates publication conduct; it does not adjudicate research misconduct on an institution's behalf.
Outcome. A decision is taken, recorded with its reasons, and communicated to everyone involved — including the person who raised the concern, so far as confidentiality allows.

Possible outcomes

No further action Request for clarification Correction or corrigendum Expression of concern Rejection before publication Retraction Referral to the author's institution Bar on future submissions

Corrections, concerns and retractions

A published article stays published. When something is wrong with it we say so on the record — we do not delete the article and we do not quietly replace the file.

CorrectionJournal error: An error introduced by the journal in production. The article is corrected and a dated notice is published and linked in both directions.
CorrigendumAuthor error: An error by the authors that does not affect the conclusions — including a corrected code or data link. The article is corrected and a notice published and linked.
Expression of ConcernUnder investigation: A credible concern is being investigated and the outcome is not yet clear. The article stays online with a linked notice, which is removed if the concern is resolved.
RetractionFindings unreliable: Findings are unreliable through error or misconduct, or the work is plagiarised, duplicated, produced through compromised review, or unethically conducted. The article stays online, watermarked RETRACTED, with a notice stating who retracted it and why.
RemovalExceptional only: Only where content is defamatory, unlawful, or presents a real risk to a person's safety or privacy. The metadata and a notice remain; the full text is withdrawn.

          Retraction notices are free to read and linked from the article. The version of record is never silently altered — every change is dated and visible. Authors are told before a notice is published and may respond, and where the journal and the authors disagree about the reason, both positions appear in the notice. Any Article Processing Charge is refunded where a retraction is attributable to the journal.
        

An honest error found by the authors themselves is not misconduct, and correcting the record is not a black mark. A researcher who writes to say that a bug changed their numbers is doing exactly what the system depends on, and will be treated accordingly.

Protection for those who report

Anyone who raises a concern in good faith is protected from retaliation by the journal, and their identity is kept confidential so far as the investigation allows. Raising a concern that turns out to be unfounded is not itself misconduct; raising one you know to be false is.

Complaints Procedure

Editors get things wrong, and processes fail. There is a route for saying so, and using it never disadvantages a future submission.

Appealing an editorial decision

Write to the editorial office within 30 days of the decision, quoting your manuscript ID. An appeal must identify one of the following:

a factual or technical error in a reviewer's report;
a misunderstanding of the manuscript by a reviewer or editor;
a review conducted outside the reviewer's competence;
a failure of the journal's stated process.

Disagreement about the significance or interest of the work is not usually grounds for appeal — that is a judgement the editors are appointed to make. Appeals are considered by an editor who took no part in the original decision, who may seek a further review. You will receive a reasoned answer within 30 days, and that decision is final.

Complaints about anything else

Editorial conduct, delays, fees or invoices, the review process, the behaviour of a reviewer, the accuracy of this website: write to the editorial office setting out what happened and what you would like done about it.

Who handles what

Complaint about a reviewer or the review processHandled by the handling editor; if unresolved, by the Editor-in-Chief, or by the Executive Editor until that appointment is made.

Complaint about an editor's conduct or a decisionHandled by the Editor-in-Chief, or by the Executive Editor until that appointment is made.

Complaint about the responsible editorReferred to the publisher, The Korean Academy, which appoints an independent person with no involvement in the matter.

Complaint about fees, invoices or waiversHandled by the Managing Editor, separately from any editorial decision.

Allegation of misconductFollows the Malpractice Policy, assessed by an editor unconnected to any party.

Concern about a published article's code or dataRaised with the corresponding author first; the editorial office takes it up where the artefact a published result depends on is missing, wrong or withdrawn.

Complaint about the publisherReferred to The Korean Academy's management, with the outcome reported to the responsible editor.

Timelines

Stage	What happens	Timeline
Acknowledgement	We confirm receipt and say who will handle the matter	5 working days
Investigation	Evidence gathered; everyone named is given a chance to respond	Up to 21 days
Outcome	A reasoned written answer, with any action to be taken	Within 30 days
Escalation	Where you remain dissatisfied, referral to the publisher	On request

Where a matter is genuinely complex we will say so and give a revised date rather than let the 30 days pass in silence.

Records and learning

Complaints and their outcomes are recorded internally. Where a complaint reveals a fault in our process, we change the process and say so — publicly, if the fault affected published work.

Where to write

Complaints and appeals: admission@thekoreanacad.com, with KJAI Complaint or KJAI Appeal in the subject line and your manuscript ID if you have one. If your complaint concerns the responsible editor, say so in the first line and it will be passed to the publisher unopened by them.