The Surge of Artificial Intelligence (AI) in Scientific Writing: Who Will Hold the Rudder, You or AI?

Recent advances in artificial intelligence (AI) and related technologies now surpass human capabilities in areas once thought to be uniquely human. AI has already outdone humans in complex reasoning tasks like chess and Go^1,2). It has been reported that some AIs have qualified for examinations for highly professional jobs^3-5). More recently, AI applications have moved beyond superficial judgments and deep learning, expanding into previously human-only domains of art and creativity.

There is no doubt regarding the expectation that scientific progress will eventually benefit humans in everyday life. However, serious concerns also exist about the easy accessibility to AI. One of the biggest concerns could be writing a scientific article. Publishing a paper in a peer-reviewed journal requires originality, creativity, logical thinking, and vital ethical considerations. It seems, however, that these criteria are no longer of the utmost importance. Generally called LLM (large language models), the current version of generative AIs can generate texts based on user prompts through the ‘natural language processing’ of ‘machine-learned’ knowledge⁶⁾. This helpful technology is just one click away from every author. Ease of accessibility also escalates concerns surrounding potential misuse. Suppose that a research group with little prior publication history suddenly submits multiple papers in a very short period. This sudden productivity may result from their extensive hard work. However, from the current perspective of an editor’s office, we must first examine whether the AI has been maliciously used. As several scientific journals have recently shared such experiences, a new set of contemporary concerns surrounding the use of AI have arisen.

Currently, most generative AI can develop research ideas from a simple user prompt, provide a theory from the concept, and write comprehensive papers with full references. At the very least, AI can provide crucial assistance in nearly all aspects of preparing a research paper. For those involved in scientific publishing, this reality poses several significant dilemmas. First, a generative AI may produce inaccurate or erroneous descriptions, known as “hallucinations.”⁷⁾ While the providers of the most current AI version claim that this phenomenon is no longer an issue, authors should be aware of the possibility. Without meticulous checks and verification by the authors, these “hallucinations” may slip through the publication process and mislead readers with incorrect information. As AI continues to improve at simulating human language, it becomes increasingly difficult to distinguish between fact and fiction and between AI and human writing⁸⁾. It raises fears that even initially flawed text could become indistinguishable from original work. Another concern is the ambiguity of applying the strict ethical standards expected of human researchers, authors, and publishers to AI tools. To illustrate this point, consider a scenario where a researcher, maliciously or inadvertently, uploads a part of their research data to an AI-powered company or program. Following the researcher’s intentions, the AI could fabricate parts of the material and create a draft manuscript. After some time, a co-worker, not recognizing where they originated, might write a scientific paper with the assistance of another AI using this dataset and the draft. This process could then lead to the fraudulent creation of a scientific article without extensive and arduous research. Worse still, if malicious intent was involved, recently popularized deepfake technology could even be employed to create or modify graphs and images in research papers^9,10). The final concern would be the ambiguity of the definition of plagiarism¹¹⁾. AI-generated text is likely to originate from the imitated content from its training database. Thus, its originality should be questioned if authors or reviewers fail to find a proper citation in every sentence. These issues compound editorial challenges. Even with specialized tools, the possibility of definitively identifying AI-generated content may diminish as technology advances. Consequently, researchers’ adherence to ethical guidelines and transparency in the research process becomes increasingly crucial. Due to these complexities, most editorial organizations hesitate to establish definitive regulations. Instead of strict rules, they state the individual stances or seek consensus.

There is widespread agreement that, rather than being a simple yes/no proposition, AI is currently used somewhere along the entire spectrum from idea generation to publication. If this is true, what is the maximum extent to which AI can be used and still define a manuscript as written by humans? How can we ascertain the originality, creativity, and integrity of scientific research?

Hip & Pelvis acknowledge that the era of AI in scientific writing has already arrived and urge all researchers to fulfill their corresponding ethical responsibilities. To make our faith manifest, we added AI-related items to the author checklist, effective August 2024. The editorial office always trusts authors’ declarations and respects their decisions. However, to accomplish our mission, we will use all available methods to verify that generative AI has not been excessively used in creating a manuscript, image, or graph. If discrepancies arise between an author’s declarations and our determinations, every effort will be made to clarify the situation. Also, the authors should take full responsibility for their statements. The current stance about the AI of the Editorial Board of Hip & Pelvis is as follows: AI cannot be recognized as an author under any circumstances. It should be used as a tool for refining papers published in Hip & Pelvis, not a magic wand for creating data or documents and realizing ideas.

Hip & Pelvis has undergone a long and winding journey since 1989. The perseverance of many researchers and former editorial members has compelled us to move forward against unexpected challenges. I believe that this journal is now on its third starting line. The first was the inauguration, and the second was the decision to expand our publication scope from domestic to international research. Now, the journal is on the line of embracing AI. Definitively, we are entering a new ocean with a new companion, not knowing what will happen after the gate closes behind us. So before you set sail, I ask you this: Who will hold the rudder of your boat, you or the AI?

Funding Statement

Funding No funding to declare.