Abstract
As I write this month’s editorial, there has been considerable discussion about the emergence of artificial intelligence (AI) in academia and concerns about the value and validity of AI-generated content. The JAMA Network recently published an update to AMA guidelines for use of AI-generated content and issues related to scientific integrity. 1 While interactive AI is interesting, there remain many concerns regarding AI-generated content. AI-generated information often lacks nuance and is routinely factually incorrect. At the current stage of development, it remains ethically problematic to rely on AI-generated content in preparation of scientific publications.
In preparing this editorial, I asked one of the publicly available generative AI tools, ChatGPT, to discuss the use of AI in academic publishing in nursing. While the response was grammatically correct, it was not a balanced discussion of AI. The generated response presented only the potential benefits of AI, but without discussion of how such benefits would be achieved and without citations to support any claims made. When I asked ChatGPT to discuss both the pros and cons of AI use in the publishing, a more balanced response was generated, but this highlights that the quality and accuracy of output from AI engines are going to be highly variable based on the engine used and the input prompts.
There have already been a few articles published in scholarly journals, including at least one nursing journal, listing ChatGPT as a co-author. In most of these cases, ChatGPT has already been removed as a co-author of these papers. The International Committee of Medical Journal Editors and the World Association of Medical Editors (WAME) have specific guidelines on what it means to be an author, which includes being accountable for the content in the published work. Because an AI system cannot take responsibility for content, nonhuman systems are not eligible to be authors of scientific work.
At this point, the Western Journal of Nursing Research discourages the use of AI for preparing manuscripts for submission to the journal. Our guidelines will be consistent with those established by SAGE Publishing (https://us.sagepub.com/en-us/nam/chatgpt-and-generative-ai), which are following recommendations by WAME 2 and largely consistent with the guidelines put out by the AMA publication network:
Nonhuman authors, including but not limited to AI bots, language models, machine learning systems, or other emerging technologies, do not qualify for authorship in WJNR.
If any AI tools or related technologies are used to generate content in the manuscript, the authors must take responsibility for the accuracy and originality of any content generated by the tool or system, and for ensuring that all information and ideas in the manuscript that are from other sources are properly attributed using appropriate citations, just as would be expected of content written by a human.
As assistive AI technology develops, authors may use AI technology for assisting with checking for grammar, spelling, punctuation, and other editing tasks when preparing manuscripts.
The use of AI tools or systems must be reported in the Acknowledgements section (for help with writing or preparing the manuscript) or the Methods section (if the use of AI or machine learning was part of the research itself).
Guidelines and policies regarding generative AI are likely to evolve as the technology evolves and our understanding grows regarding how it may be used and misused.
Editors, authors, and content creators also have concerns around generative AI using existing material to create images or other content. Copyright law (of which I am certainly not an expert) has not yet caught up with the development of these new technologies, and authors who use AI-generated content that repurposes other copyrighted material or images may place themselves at legal and ethical risk. WJNR supports the rights of the original creators of scholarly and creative work; thus WJNR discourages the use of AI-generated images and content. Authors must accept responsibility for the accuracy and integrity of any content in their manuscripts, including AI-generated content. All content originating from other sources must have citations, attribution statements, and/or permissions, as appropriate.
AI has the potential to make enormous impacts on how we do research and publish our work. It may ultimately help to improve the quality of our publications and streamline the work needed to write quality manuscripts, find reviewers, and copyedit accepted work. At this point, however, the technology is still too new and has too many flaws. Just as we would not implement new patient care technology without first thoroughly testing it for safety and efficacy, we should not rush to use new technology in generating the scientific publications that constitute the evidence base for healthcare practice. As an editor and a naturally curious person, I am looking forward to seeing how the technology develops and how it can be useful when it is ready.