Artificial Intelligence in Scholarly Publishing

Artificial Intelligence (AI) is all around us. An email titled Meet Gemini, your new AI assistant appeared in my email inbox recently. I was curious about what might be available to help me to become more efficient. I was also a little skeptical and even more cautious. Might this new assistant suggest something inappropriate, irrelevant, or even false?

Questioning the use of AI technologies is particularly relevant in scholarly publishing for authors, editors, peer reviewers, and even everyday readers. We typically ask: were the research methods sound, was the analysis done correctly, are the conclusions appropriate? What is most important to consider related to AI when interacting with the Journal of Midwifery & Women's Health (JMWH) and other scholarly journals?

First, some definitions of AI may be helpful. One expert defined AI as “algorithmic-based technologies that solve complex tasks which previously required human thinking.” Another simply referred to AI as “whatever hasn't been done yet.” In addition, the expert clarified that the intelligence part of AI includes both learning and thinking. A third expert explained AI technologies as requiring higher-level knowledge to do their work.¹

New tools referred to as generative AI or GenAI have become available in the last several years. This large language model technology refers to tools that learn from large amounts of publicly available information, including the possibility of copyrighted material, and can generate content such as human-sounding text, images, audio, and video. Numerous ethical concerns have been raised related to their use in scholarly publishing.² These are important questions to ask and answer, including topics such as intellectual property, other rights to material, privacy, and confidentiality.³ Questions are also being raised about known biases in GenAI based on the material used to train programs and the possibility of amplifying existing biases that may worsen health disparities rather than helping to make improvements. Expert humans guiding the tool development and training will be essential to prevent harm from poorly developed tools.⁴

The JMWH Editors and Associate Editors approved a new editorial policy related to AI use in 2024.⁵ The policy interprets guidance provided by thought leaders in scholarly publishing such as the World Association of Medical Editors⁶ and the JAMA Network, publisher of the AMA Style Guide, adopted by JMWH.⁷ Authors remain ultimately responsible for all information in their published work, including proper citing of sources, lack of plagiarism, and any material derived from AI tools. Any use of such tools must be acknowledged, including which tool(s) were used and how they were used. In addition, GenAI tools may not be considered authors because they are not human and therefore cannot take responsibility for the work.^5-7

Editors also have responsibilities to acknowledge the use of any AI tools. Use of these tools may include generation of content provided to authors and reviewers. Editors are responsible for using tools that aim to detect the use of GenAI content in manuscripts submitted for review.⁶ At the present time, JMWH submits all manuscripts to a similarity recognition tool in an effort to detect possible plagiarism. Authors and peer reviewers will be asked to confirm any use of AI tools in their work and declare that those were accurately and transparently described in the next update of JMWH policies.

Peer review is another area where AI tools might be used to evaluate manuscripts and write reviews. The practice of using AI to generate a peer review from a submitted manuscript is not currently permitted because it requires uploading authors’ documents to be used by the tools. This content is retained within the tool. Manuscripts submitted to journals for review for possible publication are the property of the authors. Therefore, the information cannot be shared or used in any way until the article is formally published, if that is the disposition of the manuscript.^{5, 6}

General readers, increasingly aware of disinformation produced in social media and other venues, may become more aware of possible inaccurate or false information in the scholarly literature. Readers can appropriately expect editors and publishers to be alert to misuse of tools such as writing text, generating images, and producing video material.

Ethical concerns exist about the use of AI tools; these tools must be used responsibly with human oversight and appropriate limits on their use. However, AI can be also helpful in research and publishing. Some uses include automating repetitive and time-consuming tasks such as generating and limiting literature searches to obtain the most important resources for a specific research area. Automating initial steps in the use of large data sets is another example. Related to authorship and peer review, AI tools can help editorial teams and publishers improve efficient operations, thus improving the peer review process and providing better and more timely service to authors and reviewers. AI tools can assist with revisions to manuscripts and peer reviews once they are drafted and facilitate summaries of key points of research, promoting further dissemination after articles are published. Future uses of AI might include identifying gaps in research and facilitating collaborations among scientists.⁸

Readers of JMWH can be confident that our editorial team and publisher^{3, 5} are evolving consistent with the best current thinking related to the use of AI and other technologies in scholarly publishing. We welcome hearing from authors, reviewers, and readers. JMWH will continue to observe and learn as new tools become available to both support our publishing activities and provide authors and others with guidance related to the ethical use of AI tools. Organizations that provide ethical guidance to journal editors as well as publishers can assist with advising authors, reviewers, and others in the publication process.

Human oversight and responsibility is essential in the use of AI tools. Maintaining a healthy dose of skepticism, applying the best current ethical principles and guidance, and looking for ways to confidently improve and innovate in scholarly publishing will guide our way forward. Perhaps it is time to meet Gemini.