生成式人工智慧問世帶給學術界莫大助益 ，亦對同儕審查機制帶來挑戰。本研究以文件分析法 ，首先發現在同儕審查初期階段 ，AI 科技可協助文字相似度比對、統計偵錯、找尋合適審查人 ，甚至可部分自動化。然而在第二階段 ，外部審查人接受審稿任務後 ，國外各主要研究資助機構 、學會組織與學術出版社都禁止其利用生成式 AI 審查稿件、撰寫審查意見 ，主要原因為隱私、保密 ，以及系統可能產生偏見、錯誤、訊息來源不透明等。最後當稿件回到機構或期刊 ，AI 科技或可協助綜整評語、檢查作者是否根據審查意見修改稿件 ，但最終決策仍需由人類作出。本文建議 ，台灣研究資助機構與期刊應制定 AI 使用規範 ，並強調審查者應遵守審查倫理與規範 ，以提升整體學術研究水準。

The rapid development of artificial intelligence (AI) technologies has greatly benefited the academic community. However, it has also posed significant challenges to the peer review process. This study, using document analysis, first identifies that in the initial stage of peer review, once a manuscript is submitted to a funding agency or journal, AI technologies can provide more assistance with technical checks, such as text similarity analysis, statistical error detection, format validation, and identifying suitable reviewers. Eventually, they may even automate these tasks, reducing the need for human involvement. However, in the second stage, after the external reviewers receive the manuscript for evaluation, major international research funding agencies, scholarly organizations, and publishers have prohibited human reviewers from using AI to directly review manuscripts or write review comments. The primary reasons for this are concerns over information privacy and confidentiality, as well as the limitations of current generative AI systems (e.g., biases, errors, and lack of transparency regarding income data). In the third stage, once the manuscript returns to the institution or journal, AI technologies may assist the editor in synthesizing feedback and even verifying whether the author has made revisions based on the reviewer’s suggestions; however, only humans can make the final decision on whether to accept the manuscript. This study also finds that, to date, research funding agencies and journals in Taiwan have not established specific regulations for the use of AI in peer review. Therefore, the study recommends that such regulations should be developed in the future, and reviewers should adhere to ethical standards and guidelines to provide high-quality feedback, thereby enhancing the overall academic research quality.

人工智慧 ，生成式人工智慧，同儕審查，研究資助機構，期刊出版社，審查者

Artificial intelligence, Generative AI, Peer review, Research Funding Agency, Journal publisher, Reviewer