生成式人工智能介入学位论文创作的合规性边界OACHSSCD

The involvement of Generative Artificial Intelligence(GAI)in the creation of degree theses can be categorized into four categories of behavior:low-intensity formal assistance,medium-intensity content assistance,medium-intensity content supplementation,and high-intensity content generation.The diversity of these behaviors dictates that they should not be uniformly dismissed as academic misconduct.Instead,it is necessary to delineate the boundaries of compliance,clarifying the scope of reasonable use versus academic misconduct.The determination of these compliance boundaries should follow four criteria:the originality criterion,which assesses whether the content reflects the free will and individuality of the degree candidate;the retention of critical steps criterion,which specifies that certain steps must not involve substantive intervention by GAI;and the review and verification criterion,which emphasizes the degree candidate's obligation to verify relevant content.Accordingly,low-level formal assistance behaviors such as text polishing and format optimization,moderate-level content assistance behaviors such as providing reference materials,targeted opinions,and translating or reducing the duplication of independently created content,and moderate-level content supplementation behaviors involving appropriate filling of original content,constitute reasonable use of GAI.All other GAI involvement behaviors should be classified as academic misconduct and can be precisely categorized as"academic negligence","GAI substitution of core ideas",ghostwriting,plagiarism,or falsification,based on their specific characteristics.