摘要：

【目的】分析我国学者随机生成内容(randomly generated content,RGC)撤销论文的相关特征,为有效防范因RGC撤稿提供参考。【方法】通过撤稿观察数据库(retraction watch database,RWD),检索获取我国学者RGC撤销论文数据,研究分析其时间分布、撤稿时滞、作者数量、地域与机构分布、载体与学科分布、撤稿原因等特征。【结果】2020年之后,我国RGC撤销论文量剧增,撤稿时滞高于全球水平;华东、华南和华中等地区因RGC撤稿问题较为严重,且机构分布以高校为主;文献类型主要集中在研究文章,Hindawi撤销论文量最多;RGC撤销论文的学科分为“3大技术性核心学科及其应用领域”“基础生命科学”“健康科学”3类;RGC撤销论文表现出跨学科、多元撤销原因的特征。【结论】高度警惕我国学者RGC使用规范问题,提高学术不端行为监管和发现力度,并重视RGC撤销论文地域、机构类型、载体、学科分布特征以及撤稿原因多元性,提升相关对策措施针对性、学术不端治理有效性。

关键词: 生成式人工智能, 随机生成内容, 学术不端, 撤稿, 科研诚信

Abstract:

[Purposes] To analyze the characteristics of papers withdrawn due to randomly generated content (RGC) by Chinese scholars, providing reference for domestic scientific journals, researchers, and relevant government departments to effectively address the academic misconduct issues arising from RGC abuse. [Methods] Data on RGC-related paper withdrawals by Chinese scholars were retrieved from the Retraction Watch database. The study analyzed characteristics such as temporal distribution, retraction delay, number of authors, regional and institutional distribution, publication medium and discipline distribution, and reasons for retraction. [Findings] Since 2020, the number of RGC-related paper withdrawals in China has increased sharply, with a retraction delay higher than the global average. The regions of East China, South China, and Central China are particularly affected by RGC-related retractions, with institutions involved predominantly being universities. The majority of retracted papers are research articles, with Hindawi having the highest number of retracted papers. The subjects of RGC paper retraction can be classified into three categories: the three major core technical disciplines and their application fields, basic life sciences, and health sciences. RGC-related retractions display interdisciplinary characteristics and diverse reasons for retraction. [Conclusions] It is recommended to remain highly vigilant regarding the normative use of RGC by Chinese scholars, strengthen the supervision and detection of academic misconduct, and pay close attention to the regional, institutional, publication, and disciplinary characteristics of RGC-related retractions, as well as the diversity of retraction reasons. These efforts aim to enhance the relevance of countermeasures and improve the overall effectiveness of academic misconduct governance.

Key words: Generative artificial intelligence, Randomly generated content, Academic misconduct, Retraction, Research integrity