Abstract
AbstractBackgroundChatGPT has emerged as a valuable tool for enhancing scientific writing. It is the first openly available Large Language Model (LLM) with unrestricted access to its capabilities. ChatGPT has the potential to alleviate researchers’ workload and enhance various aspects of research, from planning to execution and presentation. However, due to the rapid growth of publications and diverse opinions surrounding ChatGPT, a comprehensive review is necessary to understand its benefits, risks, and safe utilization in scientific research. This review aims to provide a comprehensive overview of the topic by extensively examining existing literature on the utilization of ChatGPT in academic research. The goal is to gain insights into the potential benefits and risks of using ChatGPT in scientific research, exploring secure and efficient methods for its application while identifying potential pitfalls to minimize negative consequences.MethodThe search was conducted in PubMed/MEDLINE, SCOPUS, and Google Scholar, yielding a total of 1279 articles and concluded on April 23rd, 2023. After full screening of titles/abstracts and removing duplicates and irrelevant articles, a total of 181 articles were included for analysis. Information collected included publication details, purposes, benefits, risks, and recommendation regarding ChatGPT’s use in scientific research.ResultsThe majority of existing literature consists of editorials expressing thoughts and concerns, followed by original research articles analyzing ChatGPT’s performance in scientific research. The most significant advantage of using ChatGPT in scientific writing is its ability to expedite the writing process, enabling researchers to draft their work more efficiently. It also proves beneficial in improving writing style and proofreading by offering suggestions for sentence structure, grammar, and overall clarity. Additional benefits identified include support in data analysis, the formulation of protocols for clinical trials, and the design of scientific studies. Concerns mainly revolve around the accuracy and superficiality of the generated content, leading to what is referred to as “hallucinations.” Researchers have also expressed concerns about the tool providing citations to nonexistent sources. Other concerns discussed include authorship and plagiarism issues, accountability, copyright considerations, potential loss of diverse writing styles, privacy and security, transparency, credibility, validity, presence of bias, and the potential impact on scientific progress, such as a decrease in groundbreaking discoveries.ConclusionChatGPT has the potential to revolutionize scientific writing as a valuable tool for researchers. However, it cannot replace human expertise and critical thinking. Researchers must exercise caution, ensuring the generated content complements their own knowledge. Ethical standards should be upheld, involving knowledgeable human researchers to avoid biases and inaccuracies. Collaboration among stakeholders and training on AI technology are essential for identifying best practices in LLMs use and maintaining scientific integrity.
Publisher
Cold Spring Harbor Laboratory
Cited by
1 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献