The Journal of Natural Law follows COPE’s Core Practices. This policy states how AI tools may permissibly be used by authors, referees, and editors.
Our policy envisions AI use as analogous to that of a research assistant or helpful colleague, and as many do not have access to either, the ubiquity of AI is a genuine gain to intellectual progress. While we value human creativity more than automated output, value does not arise exclusively from human ingenuity. Absolute bans on the use of AI are not only unenforceable but stifling to intellectual achievement. At the same time, AI output is not uniformly good: its rhetoric is routinely flat and predictable, some models are prone to make things up, and it is often wildly overconfident, among other problems. The same could be said of a research assistant. Experience paired with good judgment in a scholar can make even unreliable aids enormously useful. With this model in mind, the following policies apply to all the academic work involved at JNL.
1. General Principles
- Wording. Text is one’s own. AI tools may be used to refine original compositions but should not be the original source of any text used by author, referee, or editor.
- Degree of refinement. Minor revisions to wording, say in polishing clumsy sentences, are permitted with ordinary disclosure. Revising a whole paper in light of AI suggestions is also acceptable with disclosure. However, authorship is a property of every sentence: AI-suggested phrases may be adopted wholesale (if done sparingly), but complete sentences must remain the work of the author.
- Judgment. Arguments, assessments, and evaluations remain the sole responsibility of humans.
- AI assistance may be used to inform ideas, test arguments, generate examples, counterexamples, and objections, evaluate consistency in principle, and summarize scholarship, among others, subject to the more detailed policies below.
2. General Duties
- Verification. All are expected to verify the authenticity of AI-generated information on which they rely. Individuals are especially responsible for authenticating citations, quotations, and summaries.
- Acknowledgment. Beyond grammar and spell-checking software built into word processing software, which are exempt, all AI use must be openly disclosed by model and type of aid (e.g., brainstorming, wording improvement, research). Not disclosing AI use is a disclosure of nonuse.
- Confidentiality. Unpublished work submitted to the journal is confidential. It may only be submitted to AI tools whose terms preserve that confidentiality, by explicitly not retaining the submitted work to train models or keeping the manuscript indefinitely.
3. Authors
- AI tools cannot be authors or coauthors.
- Authors who have used AI must include an AI declaration in the text of the manuscript citing which tool was used and how, in an acknowledgment section located prior to the list of works cited. This section may (after peer review) also mention others who have helped in the paper’s improvement.
4. Referees
- A referee should read the manuscript in full and form an initial judgment on the quality of the paper without AI assistance. A recommendation may not be based on an AI summary or analysis of the paper as a substitute for reading it, and no part of the referee recommendations should be generated by AI.
- Once a referee has formed an independent assessment of the paper, AI tools may be used to help test their judgment and refine it. In particular, referees may use AI for feedback on whether they might have misread the paper, missed a response to an objection or concern, missed important literature recommendations, or to adjust tone and detail in feedback to make comments shared with the author more constructive.
- Only AI models that maintain confidentiality, as described above, may be used, and it is the responsibility of a referee not to submit a manuscript to an AI service if they have not verified that it passes this confidentiality requirement. If AI is used, referees are to remember the limits of most AI models: particularly in their defaulting to mainstream, consensus views, and in siding with the end user’s comments or criticisms.
- Referees should disclose AI use in their confidential comments to the editor, in keeping with the general acknowledgment principle above.
5. Editors
- All editorial decisions are made by the judgment of the editors.
- AI tools inform initial review of papers: in particular in checking for undisclosed AI-generated text and in an initial assessment of the paper’s engagement with the relevant literature along with in the identification of a pool of suitable referees. This occurs after the paper is read and an initial editorial judgment is made.
- After acceptance, AI is used to verify that cited sources exist in established databases (e.g., PhilPapers, CrossRef), confirm citation details and formatting, and identify writing errors before typesetting. It is also used to help identify ideal peer commentators.
- AI models used are training-disabled and only retain submitted manuscripts briefly.
- Given the evolving norms for AI use, discovered violations of this AI policy will prompt a conversation with the author or referee. Appeals of an editorial decision can be made to the journal’s associate editors. Consequences for violation may include rejection of a manuscript or retraction of a published paper.
- This policy is not retroactive and may change. Revisions to the policy will be noted by date and retained on the journal’s website (http://www.journalofnaturallaw.org/ai).
Open the policy on its own page ↗