Navigating the AI revolution: This journal’s journey continues

As powerful tools for generative AI (artificial intelligence) gain users and are increasingly implemented in science and medicine, fundamental changes are underway in the field of publishing. (1) Since its release in November 2022, OpenAI’s widely available ChatGPT, a type of neural network model trained on vast datasets, has been adopted by millions of users who are harnessing the power of generative AI for myriad tasks. As the pace of change accelerates, the integration of AI in medical publishing raises practical and ethical concerns for authors, editors, and publishers (2–6).

Traditional definitions of authorship have been upended by the advent of generative AI. For busy researchers, the initial drafting of medical papers is a time-consuming process of distilling complex research data into relatable and impactful articles. Generative AI, with its ability to generate human-like text that is contextually relevant and informative, can speed this process, formulating drafts that researchers can then refine.

The involvement of generative AI in this process prompted several authors to list ChatGPT as an author on their papers (2,7), which was promptly disputed by the International Committee of Medical Journal Editors (ICMJE). (8) Taylor & Francis Publishing, which publishes the Journal of Spinal Cord Medicine (JSCM), shares their collective opinion, stating “All authors are wholly responsible for the originality, validity, and integrity of the content of their submissions. Therefore, large language models (LLMs), generative AI, and chatbots (e.g., ChatGPT), and similar types of tools do not meet the criteria for authorship.” (9)

While generative AI has the potential to increase the volume and quality of scholarly output, it also carries risks if not utilized properly. Publishers − already dealing with plagiarism, fraudulent manuscripts, and the relentless output of papermills − see the prospect of easily fabricated submissions complicating their efforts to ensure the quality of editorial content.

A recent study by Sabel et al. underscored the current scope of the problem of “fake papers” in the biomedical literature. (10) Focusing on signs of questionable authorship, they estimated that in 2022, 28% of biomedical articles − an estimated 300,000 published articles − were faked or plagiarized. Their shocking findings made global headlines.

The researchers scanned published papers for authors with private, noninstitutional email addresses and hospital affiliations, indicators that they have determined warrant further scrutiny. While acknowledging false positives as a limitation, Sabel emphasized that their findings are disturbing. “Fake science can have a massive impact on society and on the economy,” said Sabel, editor of Restorative Neurology and Neuroscience. “It is a widespread challenge … we can no longer ignore.” (11)

Given the importance of ensuring the credibility of submissions, as editors of JSCM, it is our priority to examine current policies and procedures, and work with our publisher to upgrade safeguards, guided by the recommendations of ICMJE and the Committee on Publication Ethics (COPE) (4,5).

Our current screening of submissions involves methodology review in addition to content reviews, a procedure incorporated into JSCM’s peer review process in 2019 (12). Research papers undergo preliminary assessment by our methodology associate editors, who flag those with methodology concerns for further examination by a specialized group of peer reviewers. As part of the submission process, submissions are screened by ithenticate’s CrossRef tools for detecting signs of plagiarism. Papers with scores above a certain threshold are flagged for further review. On the publisher side, Taylor & Francis recently implemented more stringent requirements for peer review for all its journals and has updated training materials and resources for editors and peer reviewers.

Because of the widespread concerns about fraudulent papers, JSCM is adopting stricter requirements for verifying author identity and implementing guidelines for acknowledging the use of generative AI tools in the manuscript preparation.

Most importantly, we are exploring ways that AI/human synergy can help ensure the integrity of the publishing process. For example, while human judgment is essential, AI may help increase the efficiency of peer review by screening and flagging discrepancies in methodology and data reporting and improving readability. Such a hybrid model would shorten peer review and focus the time and expertise of methodology reviewers on more substantive issues and decision-making.

With the support of our Editorial Board, the Academy of Spinal Cord Injury Professionals, and Taylor & Francis Publishing, we look forward to guiding JSCM, which continues to evolve from its origin as a VA newsletter more than 45 years ago, (13) into this exciting new age of academic publishing.