隨著生成式人工智慧（Generative AI）快速普及，學術出版正進入新的人機協作階段。Elsevier近期更新期刊AI使用政策，反映國際出版界已逐漸從全面禁止AI，轉向建立透明、可追溯且符合研究倫理的治理架構。
新版政策允許編輯與審稿人將AI作為輔助工具，例如優化審稿報告文字品質、改善編輯決策信件內容，以及協助背景文獻搜尋。然而，政策明確規定，所有學術判斷、同行審查及最終決策仍應由人類專家負責。尤其禁止將未公開手稿上傳至AI平台，以避免侵犯作者智慧財產權、破壞保密義務或造成個資外洩。
對作者而言，AI已被接受作為文稿撰寫與圖像生成的輔助工具，但必須揭露使用之工具及目的。Elsevier特別強調，AI不能取代作者身份，也無法承擔學術責任。研究成果之準確性、原創性及完整性仍由作者負全責。
此政策更新顯示，國際出版界對AI的態度已逐漸成熟，不再視AI為威脅，而是視其為提高效率與促進知識生產的重要工具。然而，其核心價值仍是維護研究誠信、保護未公開研究資料，以及確保科學判斷不被演算法取代。未來AI治理模式很可能成為學術出版新常態，而透明揭露、責任歸屬與人類監督將持續是制度設計的三大基石。
政策更新背景與宗旨 (Policy Background & Purpose)
因應科技進步：隨著生成式人工智慧（Generative AI）在研究社群的加速普及，Elsevier 更新了期刊政策，旨在為作者、編輯與評審提供清晰且實用的操作指引。
維護學術誠信：政策的核心目標是在擁抱技術效率的同時，嚴格捍衛學術誠信（Research Integrity）與學術出版的信任度。
編輯與評審人員指引 (Guidelines for Editors & Reviewers)
允許的輔助範疇：編輯與評審可使用 AI 工具執行實用性任務，例如改善決策信函（Decision letters）或評審報告（Review reports）的語言流暢度與結構，或輔助進行背景文獻檢索（Background literature searches），以提升工作效率。
資訊揭露義務：評審若使用 AI 工具優化評審報告的語言或內容結構，必須主動揭露（Disclose）所使用的工具與目的。然而，單純的拼字與文法基本檢查則不需揭露。
核心禁令（嚴格保密條款）：絕對禁止將未發表的論文稿件（Unpublished manuscripts）或其任何部分複製貼上或上傳至 AI 工具中。因為投稿稿件屬機密文件（Confidential documents），此行為會嚴重侵害保密性（Confidentiality）、所有權權益（Proprietary rights）以及個人資料隱私（Data privacy）。
人類主導原則：編輯裁量權（Editorial judgement）、同行評審（Peer review）與最終決定，仍是不可轉讓的人類責任。
作者指引 (Guidelines for Authors)
文字與圖像創建：作者除了可使用 AI 輔助文稿文字準備外，新政策更進一步針對如何使用 AI 工具創建特定類型的圖表、圖像與藝術作品（Figures, images and artwork）給予明確指引。
最終問責制：作者必須對所提交內容的準確性、原創性（Originality）與誠信負起完全責任。AI 不能取代作者身份（Authorship）、批判性思考（Critical thinking）或學術責任（Accountability）。
獨立揭露聲明：作者在投稿時，必須透過獨立的「AI 聲明書（AI declaration statement）」主動揭露相關的 AI 使用情況（包含工具名稱與目的）。基本的拼字、文法與標點符號檢查免予宣告。
圖像特別標記：若圖像（Images）的創建使用了 AI，作者除了在整篇論文的 AI 揭露聲明中提及外，還必須在該圖像的圖說（Caption）中明確標註。
Updated Elsevier policy on generative AI for journals
Dear Editor,
As you may have seen, Elsevier has updated its policy on generative AI for journals. The updated policy provides clear guidance on how AI tools may be used by authors in manuscript preparation, including image creation, and by editors and reviewers in their editorial workflows.
https://www.elsevier.com/about/policies-and-standards/generative-ai-policies-for-journals
As AI adoption accelerates across the research community and publisher policies continue to evolve, we update our policies to keep them clear, practical, and aligned with the broader landscape while maintaining our standards for research integrity and confidentiality.
What this means for editors and reviewers
The updated policy gives editors and reviewers guidance on how generative AI tools can be used to support parts of their work. AI tools can help with practical tasks such as improving the language or structure of decision letters and review reports, and supporting background literature searches, helping make some parts of the process more efficient.
When reviewers use AI to improve the language or content structure of their review report, they should disclose this, including the tool used and purpose. Basic checks for spelling and grammar do not need to be disclosed. For editors, journals will inform authors that AI tools may be used in a supportive capacity via the Guide for Authors and EM letters.
What stays essential
Editorial judgement, peer review and decisions remain human responsibilities. Editors and reviewers must not upload or copy and paste unpublished manuscripts, or parts of manuscripts, into AI tools. That is because submitted manuscripts are confidential documents, and doing so may compromise confidentiality, proprietary rights, or data privacy where personal information is involved.
What this means for authors
Authors may already use private AI tools to support textual manuscript preparation, and the updated policy gives guidance on how to use AI tools to create certain types of figures, images and artwork
Authors remain fully responsible for the accuracy, originality, and integrity of everything they submit. AI can support the process, but it does not replace authorship, critical thinking, or accountability.
Authors must disclose relevant AI use in manuscript preparation through a separate AI declaration statement at submission, identifying the private tool used and the purpose of use. Basic spelling, grammar and punctuation checks do not need to be declared.
Where AI is used for images, authors should also disclose that use in the image caption and in the article’s general AI disclosure statement.
The aim of this update is to give clearer guidance as AI tools evolve, while protecting trust in scholarly publishing. Elsevier will continue to keep the policy under review as practice and technology develop.
Publisher
No comments:
Post a Comment