Reporting Use of AI in Research and Scholarly Publication—JAMA Network Guidance

Reports on the use of artificial intelligence (AI) and machine learning, including large language models (LLM), in medical research have intensified in the last year. Although machine learning research began 70 years ago with the conceptual development of artificial neural network algorithms,¹ AI research and use in clinical practice and health care are relatively recent advances. Throughout these developments, JAMA has sought to define the broad scope of discovery and innovation in medical applications of AI and to address potential challenges in its implementation. The journal’s 2016 publication of a study of deep learning algorithms for the detection of diabetic retinopathy from fundal photographs is a useful example.² The study represented a novel tool that could enable large-scale screening for a key vision-threatening disorder across the world. However, the accompanying Editorial³ highlighted important challenges, spanning the need for broader patient representativeness, the investment necessary in validating the model in the context of its deployment and subsequent implementation, and whether clinicians would entrust decision-making to AI tools. The Editorial also called attention to concerns regarding AI eventually replacing humans in clinical systems. Although published before the recent AI boom,⁴ this study and the comments about it augured many of today’s promises and concerns regarding AI in clinical research and practice.^5–11

Over the past decade, advances in AI have enabled many innovations that facilitate rapid research and resulted in tools for disease prediction, diagnosis, and prognostication. During the last year, JAMA provided guidance to authors and peer reviewers on the transparent, appropriate, and accountable use of AI.^12,13 Herein, we provide more detailed recommendations for authors and researchers. Editors and readers rely on authors to report their use of AI and to be fully accountable for such use. Transparency is crucial in employing AI technologies given its rapid development and yet-unknown functionalities. The guidance below is general and based on some common problems we have seen in submitted manuscripts; reporting guidelines will provide more specific guidance. This information has been added to the Instructions for Authors for JAMA and the JAMA Network journals.^{14 …More}