AI Tool Policy

This policy governs the use of generative artificial-intelligence and AI-assisted tools by authors, reviewers and editors of the Journal of Constitutional Change. It is reviewed periodically as the technology and the relevant guidance change, and was last revised in [bulan tahun].

The governing principle

Responsibility for the content of a manuscript cannot be delegated to a tool. An author who submits to JCC warrants that every proposition in the manuscript — and in particular every proposition of foreign constitutional law — has been verified by a human being against a source that human being has personally consulted. Whether a tool was used along the way matters less than whether this warranty is true.

Artificial intelligence cannot be an author

AI systems do not satisfy the Journal's authorship criteria: they cannot approve a final manuscript, cannot be accountable for its accuracy, and cannot respond to allegations of error. AI tools must not be listed as authors or co-authors, and must not be credited in the author line, the byline or the acknowledgements as if they were contributors.

Tiered disclosure

Every submission includes an AI Use Statement, placed after the acknowledgements, identifying the tier of use. A statement is required even where no tools were used; silence is not an option, because a reader cannot distinguish silence from concealment.

Tier

Use

Required disclosure

Tier 0

No generative AI tools were used at any stage.

"No generative artificial-intelligence tools were used in the preparation of this manuscript."

Tier 1

Language and presentation only: grammar and style correction, translation of the author's own prose, reference formatting.

One sentence naming the tool, its version and the use. Example: "[tool name and version] was used to correct grammar and style in the author's own English prose."

Tier 2

Analytic assistance: literature searching, summarising sources the author has read, transcription, coding or statistical scripts, structuring a draft the author wrote.

Tool, version and a specific description of each use, together with a statement of how the output was verified.

Tier 3

Generative contribution: text generated by a tool is retained, in whole or in part, in the scholarly content of the manuscript.

Full account of the passages concerned and the prompts used, supplied to the editors on request. The editors will assess whether the authorship criteria are still satisfied. Tier 3 use is permitted only where declared; undeclared Tier 3 use is treated as a breach of publication ethics.

 

Prohibited uses

The following are prohibited in all tiers, because each produces material that appears authoritative and cannot be relied upon. Constitutional scholarship is unusually exposed to this risk: a fabricated article number, a plausible but non-existent decision, or a subtly wrong translation of an entrenchment clause can pass unnoticed by readers who do not work in the jurisdiction concerned, and can then be cited onward indefinitely.

  • Generating, paraphrasing or reconstructing the text of a constitutional provision, statute, judgment, drafting record or treaty. Primary materials are quoted from the source, not from a model.
  • Translating primary legal materials without verification against the original text by a person competent in the source language. Where an AI tool assisted a translation, this is stated in the footnote and the verifier is named.
  • Generating citations, bibliographies or quotations. Every reference must correspond to a source the author has consulted.
  • Generating, augmenting or imputing empirical data, interview material, or images presented as evidence.
  • Producing peer-review reports, editorial decisions, or correspondence presented as the work of a named reviewer or editor.

Reviewers

Reviewers must not enter any part of a manuscript — including the title, abstract or a single paragraph — into a generative AI system, whether to summarise it, assess it, or improve their own report. Manuscripts under review are confidential, unpublished work, and uploading them to a third-party service is a disclosure of that work regardless of any assurance the service offers. A reviewer who is unable to work within this restriction should decline the invitation; no explanation is required and the editors will not regard it as a lack of cooperation.

Editors

Editorial decisions are made by human editors on the basis of human reviews. AI tools may be used in the editorial office only for administrative functions that do not involve evaluative judgement — similarity screening, reference checking, image integrity checking, and metadata preparation — and every output of such a tool is confirmed by an editor before it has any consequence for an author.

How suspected undisclosed use is handled

The Journal does not use, and does not accept the output of, AI-detection software. Such tools produce false positives at rates that are unacceptable when the accusation is misconduct, and they disadvantage authors who write in English as an additional language — precisely the authors JCC exists to publish.

Concerns are instead pursued through verification. Where the editors doubt the provenance of material, they ask the author to produce the sources relied upon: the decision at the paragraph cited, the constitutional text in the original language, the data behind a figure. An author who can produce the sources has answered the concern. An author who cannot has a problem that exists whether or not a tool was involved, and the matter proceeds under the Publication Ethics policy.

Why the Journal does not simply ban these tools

A prohibition would fall hardest on scholars writing in a second or third language and on those without access to professional editing or large research libraries — a significant share of the constitutional-change scholarship the Journal wishes to publish. The Journal's position is that the appropriate response to unreliable tools is not exclusion but verification, applied to everyone equally.