UMMS Artificial Intelligence

A review of policies on the use of artificial intelligence or automation in manuscript submissions from several top journals publishing articles in medicine reveals a few common rules and requirements:

Disclosure: generally, use should be disclosed. Some specifics instances:

• Use of artificial intelligence must be disclosed at the time of manuscript submission.
• Most policies require the name of the tool and a description of content provided by the tool or a description of how the tool was used.
• Some policies require this disclosure in forms submitted with the publication and other policies stipulate that disclosures must also be made in the manuscript itself, typically in the methods section.
• The Lancet requires that the description of the tool and its use be sufficient for replication.
• The Lancet requires the disclosure of AI that was used to improve readability.

Allowed uses not requiring disclosure

• Many policies allow the use of grammar and spelling checkers and reference software and do not require that their use be disclosed.

Authorship

• Most policies specify that large language models (LLMs), like ChatGPT, do not qualify for authorship.
• Some policies further indicate that only humans can be authors.
• Most state that authors must be fully responsible for all content, including content produced by AI.
• BMJ, JAMA, and Science all require that authors using AI consider bias. 

Several journal policies reference guidance developed by the International Committee of Medial Journal Editors (ICMJE) and the Committee on Publication Ethics (COPE).

Policies and guidance on the use of AI will inevitably continue to evolve as the tools change and mature, therefore, periodic review of journal policies for changes will be necessary. Currently, many journals do not describe any enforcement mechanisms for their AI policies, but tools to detect AI generated content are already in use, including a tool for Springer Nature. 

Below are links to the policies in selected journals publishing articles in medicine. Click on the journal name to read the policy in full.  

Academic Medicine

Biomed Central Journals

BMJ

JAMA

The Lancet

Nature

NEJM

PLoS ONE

Science

To find policies in journals not listed here, check the journal website for variously named submission guidelines (e.g., author guidelines, instructions for authors, publishing practices, editorial policies).

There are several examples of guidance on disclosing the use of AI and automation:

• The AID Framework, inspired by the Contributor Role Taxonomy (CRediT) https://credit.niso.org/, provides guidance on disclosing AI use in research, writing, and in education settings. Weaver, K. D. (2024). The Artificial Intelligence Disclosure (AID) Framework: An Introduction. arXiv preprint arXiv:2408.01904.

• Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) includes (Item 8) the statement: "Specify the methods used to decide whether a study met the inclusion criteria of the review, including how many reviewers screened each record and each report retrieved, whether they worked independently, and if applicable, details of automation tools used in the process." Find the key PRISMA documents at prisma-statement.org

• Reporting guideline for the early stage clinical evaluation of decision support systems driven by artificial intelligence (DECIDE-AI): includes 17 AI specific reporting items with 28 subitems. The paper describing the guidance development can be found at  https://www-bmj-com.proxy.lib.umich.edu/content/377/bmj-2022-070904. The Explanation and Elaboration document  and checklist can be found in the Supplementary appendices 1 and 2. 

• Checklist for Artificial Intelligence in Medical Imaging (CLAIM): The paper describing the development of the checklist, including the checklist items can be found at: https://doi-org.proxy.lib.umich.edu/10.1148/ryai.2020200029