Artificial intelligence and the creation of scientific papers

The field of artificial intelligence continues to develop. The term artificial intelligence (AI) has recently become a buzzword in many circles, and there are occasions when the term AI is used incorrectly. Alan Turing, considered by most the father of AI, initiated this field when he wrote his 1950 essay Computing Machinery and Intelligence,9Turing A.M. Computing machinery and intelligence.Mind. 1950; 49: 433-460Crossref Google Scholar attempting to answer the question “Can machines think?” Those interested in learning more about Turing may enjoy the movie “The Imitation Game.”10Tyldum M. The Imitation Game. Black Bear Pictures, Bristol Automotive, and Orange Corp, 2014Google Scholar The term AI was coined by John McCarthy in 1955, who defined AI as “the science and engineering of making intelligent machines.”4McCarthy J. Programs with common sense. Stanford University, Computer Science Department, 1959Google Scholar Currently, AI is defined as “the theory and development of machines/computer systems able to perform tasks that require human intelligence.”3Joiner I.A. Emerging library technologies. Chandos Publishing, 2018Google Scholar Human intelligence allows learning, processing language, processing visual information, recognizing patterns, and performing several other tasks. Corresponding AI areas include machine learning (which includes deep learning [supervised or unsupervised] and neural networks), natural language processing, computer vision, and other. Some AI products are already present in our daily lives. For example, computer systems analyze the content that viewers consume in streaming platforms such as Netflix, learn what each user likes,6NetflixHow Netflix’s recommendations system works.https://help.netflix.com/en/node/100639Date accessed: February 5, 2023Google Scholar and offer that viewer additional content that he or she will likely enjoy consuming. AI has the potential to enhance our lives by providing accuracy and precision, speed, and lack of fatigue that may be superior to human performance, but AI also has the potential for errors and unfortunate outcomes. AI is becoming more common in health care, medicine, research, and science. The objectives of the Journal of Shoulder and Elbow Surgery (JSES) are to enhance the professional study and practice of shoulder and elbow surgery, to act as a stimulant to research by providing a forum for discussion of new scientific advances, and to further international cooperation among shoulder and elbow societies by serving as an official publication for recognized societies. Consequently, JSES and our Family of Journals need to consider the impact of AI in research and paper submissions. AI has the potential to provide powerful research tools that have already been utilized in the field of shoulder and elbow surgery. For those interested in the potential for AI to allow us to identify causation, and not just association, we would recommend The Book of Why by Pearl and Mackenzie.7Pearl J. Mackenzie D. et al.The book of why: the new science of cause and effect. Basic Books, 2018Google Scholar Undoubtedly, well-utilized AI research tools will enhance our ability to gain understanding in shoulder and elbow surgery, and JSES will continue to process AI research submissions through our peer-review process. A different use of AI in the creation of scientific manuscripts has been in the spotlight recently, especially in relation to a chatbot launched in November 2022 called Chat Generative Pre-Trained Transformer (ChatGPT).1ChatGPT.https://openai.com/blog/chatgpt/Date accessed: February 5, 2023Google Scholar This AI chatbot can provide detailed responses and answers across multiple domains of knowledge. Recently, ChatGPT was reported to pass the United States Medical Licensing Exam.5Medscape A.I. Bot ChatGPT Passes US Medical Licensing Exams Without Cramming – Unlike Students.https://www.medscape.com/viewarticle/987549Date accessed: February 5, 2023Google Scholar Others have reported that ChatGPT can write believable scientific abstracts and scientific papers, and in fact some scientific submissions have listed ChatGPT as a coauthor. It is important to note that this particular program produces language based on algorithms that incorporate the frequency that words are associated with each other. It does NOT generate ideas or interpret the ideas that may develop from assembling words together. As a result, the information produced may not be accurate or reliable. The International Conference on Machine Learning recently announced a policy stating that submissions produced entirely with large-scale language model (LLM), such as ChatGPT, are prohibited, raising questions such as “How to define an author?” or “Where to draw the line between writing and editing?”2Fortieth International Conference on Machine Learning.https://icml.cc/Date accessed: February 5, 2023Google Scholar In response to the increased use of generative AI and AI-assisted technologies by authors of scientific submissions, JSES and its Family of Journals would like to adopt a policy that was created with particular focus in ensuring the integrity of published peer-reviewed research. This policy has been reviewed and approved by Elsevier and all JSES Editors. The complete text of the policy will be made available online to prospective authors, but the essence of the policy is summarized in the following bullet points:•When writing a scientific article, AI and AI-assisted technologies can only be used to improve the readability and language of the submission, but they cannot replace key research tasks.•AI and AI-assisted technologies must always be used with human oversight and control.•The use of AI and AI-assisted technologies for manuscript writing and editing must be disclosed at the time of submission.•AI and AI-assisted technologies cannot be listed as an author or coauthor. Further, JSES and all journals within the JSES Family of Journals will request additional information for any author(s) using ChatGPT or other similar AI technologies now available, or to be invented in the future. This will include that author(s) must document that these technologies were used and on which sections of the manuscript they were used. JSES will also reserve the right to reject any article in which such technology has been used in more than a “small” portion of the article, though we cannot yet define “small” in this regard. Recently, many research articles have shown that it is not always possible, and in fact, is very difficult to determine, whether documents were produced by ChatGPT rather than by a human. JSES editors realize we are no better than others, and will also have this difficulty. Paraphrasing Wilford Brimley at the end of “Absence of Malice,” “Now, you know and I know that we can’t tell you what to [write] or what not to. We hope [that medical authors] will act responsibly [with regards to AI] but when you don’t there ain’t a hell of a lot anybody can do about it.”8Pollock S. Absence of Malice. Mirage Enterprises and Colombia Pictures, 1981Google Scholar At least not yet, but we must trust you in this regard. This policy will need to be modified as the field of AI continues to evolve, but at a minimum it provides a framework of reference that will help preserve integrity and honesty for authors contributing to the field of shoulder and elbow surgery through JSES and the JSES Family of Journals.