Navigating Scientific Peer Review with ChatGPT: Ally or Adversary?

Abstract

ChatGPT (Generative Pre-trained Transformer), a breakthrough innovation by OpenAI) is being touted as a revolutionary tool with immense potential in an array of medical and pharmaceutical research and scientific peer review (SPR) processes. 1,2Studies revealed that ChatGPT might act as a complementary tool to the human SPR and aid in expediting the process, reduce reviewer fatigue, and shorten publication timelines. 3,4Of note, ChatGPT displayed remarkable competence in providing shrewd feedback, detecting methodological defects, and measuring the article's impact on the advancement of the respective field, all with a fair inter-rater agreement. 5n the other side, a recent article by Liang et al. raised an alarm regarding the perils of using ChatGPT in the peer review process.The study found remarkable alteration using ChatGPT in nearly 17% of the peer-review reports. 6The researchers have analyzed about 146,000 peer reviews submitted to the AI conferences (pre-and post-launch of ChatGPT) and found a remarkable upsurge in the use of certain buzzword adjectives like versatile, meticulous, intricate, etc. (the telltale signs of ChatGPT-written text) in the review reports.ChatGPT has limited utility in the SPR due to a lack of transparency in training data and decision-making process, issues with the reproducibility of review reports, inability to justify the recommendations, potential biases, and AI hallucinations (generation of fake/non-existing references for writing and reviews).Besides, lack of contextual expertise, missing human connect (iterative fine-tuning, personal interaction and collaboration between reviewers and

Ähnliche Arbeiten