Yale J Biol Med. 2023 Sep;96(3): 415-420
The increasing volume of research submissions to academic journals poses a significant challenge for traditional peer-review processes. To address this issue, this study explores the potential of employing ChatGPT, an advanced large language model (LLM), developed by OpenAI, as an artificial intelligence (AI) reviewer for academic journals. By leveraging the vast knowledge and natural language processing capabilities of ChatGPT, we hypothesize it may be possible to enhance the efficiency, consistency, and quality of the peer-review process. This research investigated key aspects of integrating ChatGPT into the journal review workflow. We compared the critical analysis of ChatGPT, acting as an AI reviewer, to human reviews for a single published article. Our methodological framework involved subjecting ChatGPT to an intricate examination, wherein its evaluative acumen was juxtaposed against human-authored reviews of a singular published article. As this is a feasibility study, one article was reviewed, which was a case report on scurvy. The entire article was used as an input into ChatGPT and commanded it to "Please perform a review of the following article and give points for revision." Since this was a case report with a limited word count the entire article could fit in one chat box. The output by ChatGPT was then compared with the comments by human reviewers. Key performance metrics, including precision and overall agreement, were judiciously and subjectively measured to portray the efficacy of ChatGPT as an AI reviewer in comparison to its human counterparts. The outcomes of this rigorous analysis unveiled compelling evidence regarding ChatGPT's performance as an AI reviewer. We demonstrated that ChatGPT's critical analyses aligned with those of human reviewers, as evidenced by the inter-rater agreement. Notably, ChatGPT exhibited commendable capability in identifying methodological flaws, articulating insightful feedback on theoretical frameworks, and gauging the overall contribution of the articles to their respective fields. While the integration of ChatGPT showcased immense promise, certain challenges and caveats surfaced. For example, ambiguities might present with complex research articles, leading to nuanced discrepancies between AI and human reviews. Also figures and images cannot be reviewed by ChatGPT. Lengthy articles need to be reviewed in parts by ChatGPT as the entire article will not fit in one chat/response. The benefits consist of reduction in time needed by journals to review the articles submitted to them, as well as an AI assistant to give a different perspective about the research papers other than the human reviewers. In conclusion, this research contributes a groundbreaking foundation for incorporating ChatGPT into the pantheon of journal reviewers. The delineated guidelines distill key insights into operationalizing ChatGPT as a proficient reviewer within academic journal frameworks, paving the way for a more efficient and insightful review process.
Keywords: chatGPT; journal review; review