AI Usage Guidelines

Hume Studies aims to publish original work of the highest scholarly standards on any aspect of the thought of David Hume. It permits thoughtful and responsible AI use consistent with this aim. This document identifies some of the major risks users of AI may incur, and lays out policies governing how the journal's Authors, peer Reviewers and Editors may use AI.

This journal's AI Usage Policy is reviewed annually by the Hume Society. All parties to Hume Studies should check regularly for updates.

In addition to their utility, there are significant risks attendant on the use of the current generation of AI tools ranging from loss of rights to one’s own work, to inaccuracy and bias, to participation in exploitative labor practices, to adverse environmental impacts. Major risks for researchers include the following:

·         Absence of Responsibility: AI tools are not authors, as they have no responsibility for the content they generate.

·         Inaccuracy: Current AI tools can give incorrect answers, and do not always generate outputs that correspond to what their users intended.

·         Source Bias: Unless linked to institutional data bases (i.e., to public or private data bases other than those available on the web), AI tools do not have access to the significant body of academic journals that exist behind paywalls. This reliance on open source and public domain material poses an added bias risk. Users must be aware of these risks and take measures to mitigate them when possible.

·         Source Erasure, Fabrication, and "Hallucination": Unlike traditional search engines, which quote referenced material, current AI tools generate original responses to queries by scanning the web for (normally undisclosed) relevant material and then applying statistical scripts to mold that material into a response that has a high probability of addressing the query. While the response may not quote any human author(s), the concept is likely to be a variant (possibly a seriously incorrect one) on what others have said. Where human authors are quoted, no such individual or no such publication by that individual may exist, or if they do, they may never have said any such thing.

·         Violations of Confidentiality & Data Privacy: AI tools can seize ownership of material uploaded to them. Uploading your own or another's work to AI tools may breach the confidentiality of others and void your own expectations of confidentiality and/or data privacy.

·         Violations of Intellectual Property: AI tools can seize ownership of material uploaded to them. Where the material is the original work of an author the tool could use that material for training (influencing its responses to subsequent queries), apply very liberal open access licencing (allowing unattributed use) to the material, or prohibit others (including Hume Studies) from publishing it because it breaches the intellectual property rights of another and/or voids that of the author.

I. Authors
Authors must always disclose any use they make of AI tools, indicating purpose, impact on key arguments or conclusions, and the human oversight strategy used to control AI’s results. Disclosures must further attest that AI was not used to generate core content. Disclosures must be included in the AI Disclosure note.

When preparing and submitting their manuscripts, authors are responsible for ensuring that their use (if any) of AI tools does not compromise the originality of their work or leave that of others unacknowledged. Authors are advised to review the terms and conditions, terms of use, or any other conditions or licenses associated with the AI tool they intend to use.

Authors must exercise due diligence over determining whether anything discovered using AI was drawn from a human author, referencing that source, verifying that the source has been represented accurately, and omitting or correcting material that cannot be so sourced or verified. Authors also need to ensure that the AI tool they employ is appropriate and robust for their proposed use, and that the terms applicable to that tool provide sufficient safeguards and protections, particularly regarding intellectual property rights, confidentiality, and security.

AI tools may not be cited or listed as authors.

Acceptable uses of AI tools are limited to:
•      making grammatical or spelling corrections to originally written text
•      editing occasional sentences in the manuscript, such as by following recommendations to improve accessibility, clarity or readability, make idiomatic word choices or rephrase awkward sentences, or reduce word count 
•      formatting data tables
•      formatting references or citations
•      brainstorming ideas (provided the subsequent work meets the journal's Mandate and Scope for originality and importance)
•      conducting an initial literature discovery (provided that the citations are manually verified by the author)
•      translating occasional isolated sentences used in excerpts or quotations, except as noted below.

Impermissible uses include but are not limited to:
•      translating any text that, if left in the original language, could not be cited without the author’s or publisher’s permission
•      generating references or citations
•      generating or modifying images
•      generating or collecting data
•      generating interpretation, analysis, theses, or other core content.

 In the event of ambiguity or uncertainty over AI usage, the author(s) should contact the Editors.

 Authors are responsible for making sure their manuscripts comply with this policy, but Hume Studies maintains the right to reject manuscripts or retract articles deemed to be in violation of it. This right of rejection/retraction does not relieve the author of liability for breach of the policy and responsibility to Hume Studies and the public.

 

II. Authors of Book Reviews
Authors of book reviews are bound by the guidelines for authors.

In addition, authors of book reviews must neither utilize AI tools to generate summaries of the work(s) being reviewed nor consult AI-generated summaries of such work(s).

 

III. Reviewers
Reviewers must not upload files, images or information from unpublished manuscripts into AI tools, be it in whole or in part. See the risks about confidentiality and intellectual property rights identified under "Risks."

Reviewers must not upload their own reports to AI tools, even for purposes of translation or improving readability, as this may disclose confidential information about the manuscript and/or the authors.

Reviewers must not use AI tools to generate their reports.

 

IV. Editors

The Editors shall abide by the same disclosure principles mandated for reviewers. They are not at this time making any use of AI tools for the processing, review, or editing of submissions. Note that Hume Studies, like other leading academic journals, relies on machine-readable metadata to enhance the discoverability of its content. In addition, all volumes prior to the five most recent are publicly available online.