Generative AI, and Scholarly Manuscripts: WAME Recommendations on Chatbots and Generative Artificial Intelligence in Relation to Scholarly Publications

TL;DR

This revised WAME statement provides recommendations for editors, authors, and reviewers on chatbots and generative AI in scholarly publications. It adds a new recommendation on editor/reviewer disclosure and emphasizes human authorship, transparency, responsibility, confidentiality, and detection tools.

Key findings

• Chatbots cannot be authors: they do not meet ICMJE authorship criteria, cannot give final approval, cannot be accountable, cannot understand conflict-of-interest statements, and have no affiliation independent of their developers.
• Authors must be transparent about chatbot use: Recommendation 2.1 says chatbot/AI use to draft new text should be noted in acknowledgments, with all prompts specified; Recommendation 2.2 says analytical work, results reporting, tables/figures, or code generation should be stated in the Abstract and Methods, including full prompt, time/date of query, AI tool, and version.
• Authors are responsible for chatbot-provided material, including accuracy, absence of plagiarism, and appropriate attribution; they must identify the chatbot and specific prompt and mitigate plagiarism risk, provide balanced views, and ensure references are accurate.
• Editors and peer reviewers should specify any chatbot use in manuscript evaluation, reviews, and correspondence; they are responsible for generated content and citations, and should know that supplying a manuscript to a chatbot breaches confidentiality.
• Editors need appropriate tools to detect AI-generated or altered content, made available regardless of ability to pay, to help ensure integrity of health-care information and reduce adverse health outcomes.

Methods

• This is a revised consensus/recommendation statement from WAME, updating the earlier “WAME Recommendations on ChatGPT and Chatbots in Relation to Scholarly Publications” (January 20, 2023).
• It adds a new recommendation (#4) to the four original principal recommendations and tailors recommendations to three uses of chatbots in scholarly publishing: simple word-processing tasks, generation of ideas and text, and substantive research.
• No empirical design, sample, data collection, or formal analysis is reported in the paper.

Limitations

• The paper states that the field is rapidly evolving and that the recommendations will continue to be modified as software and applications develop.
• No empirical validation, data, or formal evidence synthesis is reported; this is evident from the text.
• No details are provided on how the recommendations were developed, reviewed, or agreed upon beyond their issuance as WAME recommendations; this is evident from the text.