AI Policy

This page governs the use of generative artificial intelligence in preparing, reviewing, and handling manuscripts submitted to Real Coster. It does not restrict artificial intelligence as a subject of study: accounts of community programmes that teach, deploy, or examine such tools are within scope.

Data about communities must not leave the research

This is the restriction that matters most in this journal, and it is absolute.

Community service generates material about identifiable people: interview recordings and transcripts, household survey responses, attendance registers, health and income data, photographs, field notes naming individuals and places. None of it may be submitted to a generative artificial intelligence tool for transcription, translation, coding, summarising, or any other purpose.

Participants consented to the engagement and to its publication. They did not consent to having their words and circumstances transmitted to a commercial service, retained on servers abroad, and possibly used to train a model. Doing so breaches that consent whether or not the material is later published, and is treated under the Research Misconduct page.

Where such processing is unavoidable, the material is fully de-identified first, and the manuscript states what was processed, with which tool, and how de-identification was carried out.

For authors

Permitted, subject to disclosure

  • Improving the clarity, grammar, or readability of text the authors wrote
  • Translating the authors’ own text between Bahasa Indonesia and English
  • Formatting references
  • Supporting the search for literature, provided every source cited has been located, read, and verified in its original form

Not permitted

  • Generating substantial parts of the manuscript and presenting them as the authors’ own
  • Producing, altering, or extending field data, interview responses, attendance figures, or outcome measures
  • Generating references, citations, or quotations, including from Scripture. Such tools produce plausible sources that do not exist
  • Generating photographs or images presented as documentation of an activity, a participant, or a place
  • Producing the account of what changed. The claim of impact is the core of a community service manuscript and must come from the evidence, not from a model’s sense of how such accounts usually read

Authors remain fully responsible for everything bearing their names, including material prepared with assistance. Responsibility cannot pass to a tool, its developer, or any service using it.

Disclosure

Where a tool has been used in any permitted capacity, the authors say so in the declarations before the references, naming the tool and version and describing the purpose in a sentence or two. A manuscript prepared without such assistance states that rather than remaining silent.

Generative output is not a citable source, since a reader cannot retrieve or verify it. Such use is recorded in the declarations and does not appear in the list of references.

Disclosure made in good faith is never a ground for rejection. Undisclosed use discovered later is a breach of publication ethics.

For reviewers

A manuscript under review is confidential and the intellectual property of its authors, and in this journal it also carries information about communities who are not party to the review. It may not be uploaded to a generative tool for any purpose, including summarising, language checking, and translation, since each requires the text to be transmitted.

The evaluation itself may not be delegated. A review is an expert judgement for which the reviewer is personally accountable. Reviewers may use such tools to improve the language of a report they wrote themselves, provided it contains no identifying or confidential content at that point. Any such use is disclosed to the editorial board.

For editors

Editors are bound by the same restriction on confidentiality. Editorial decisions are made by people and are not generated or determined by artificial intelligence. Such tools may be used for administrative tasks that transmit no manuscript content. Similarity screening is carried out through the service named on the Plagiarism page.

Significant use of such tools in the editorial process is disclosed.

Detection

The journal does not treat automated detection as sufficient evidence of undisclosed use. Such tools produce false results in both directions and disadvantage authors writing in a second language, which in this journal includes many practitioner authors writing in English.

Where the editors have reason to question how a manuscript was prepared, they put the matter to the corresponding author and invite an explanation, with supporting material such as drafts, field notes, or original sources, before reaching any conclusion. The procedure is the one on the Research Misconduct page.

Enforcement

The journal may ask for an account of how a tool was used, examine any case of undisclosed or improper use, and issue a correction or retraction where misuse is established after publication.

Review of this policy

The editorial board reviews this page annually and announces any revision through the Announcements section of this website.