Abstract:

Purposes This study aims to analyze the usage norms of generative artificial intelligence （GenAI） in academic journals of the social sciences in China， offering guidance for the news and publishing industry to establish academic publishing standards that prevent unethical practices caused by GenAI tools. It also provides references for academic journals to formulate and improve GenAI usage norms， thereby advancing the high-quality development of academic journals. Methods We investigate the usage norms of GenAI in academic publishing among the 660 social science journals indexed in the CSSCI （Chinese Social Sciences Citation Index， from 2023 to 2024）. By employing the BERTopic model for thematic identification， it aims to understand the characteristics of GenAI usage norms and their focal areas in academic journals. Findings The findings reveal that only 7.58% of social science journals in China have established GenAI usage norms in academic publishing， with significant differences across disciplines. The publication of these norms by journals lacks standardization， resulting in authors， editors， and reviewers often being unable to access timely requirements regarding GenAI usage. Moreover， existing norms impose requirements more clearly on authors than on editors and reviewers， and the specific provisions of these policies have not yet reached a consensus. Conclusions To address these issues， social science journals should develop and refine GenAI usage norms in academic publishing across four dimensions： journals， authors， editors， and reviewers. At the journal level， measures include promptly establishing and improving norms based on real needs， publishing norm content through multiple channels， enhancing training for reviewers and editors in identifying GenAI-generated content， actively developing or adopting AIGC detection services， and establishing a tiered penalty system for violations. For authors， principles such as reasonable use， prohibited use， accountability， disclosure requirements， and penalties for violations should be clarified. For editors， principles including prohibiting GenAI involvement in decision-making， AIGC detection， and intellectual property protection should be established. For reviewers， principles like confidentiality， independence， transparency， and oversight should be clearly defined.

Key words: GenAI usage guidelines, BERTopic model, AIGC detection, Publishing ethics guidelines, AI ethics

摘要：

目的 分析中国社会科学领域学术期刊发布的生成式人工智能（GenAI）使用规范，为新闻出版行业制定学术出版标准防范使用GenAI工具导致的失信行为和学术期刊制定与完善GenAI使用规范提供参考，推动学术期刊的高质量发展。 方法 调研CSSCI（2023—2024年）收录的660种社会科学领域学术期刊中关于学术出版中GenAI的使用规范，基于BERTopic模型对其规范进行主题识别，了解学术期刊制定的GenAI使用规范特点与关注内容。 结果 当前社会科学领域仅有7.58%的学术期刊发布了学术出版中GenAI使用规范，且各学科之间存在较大差异；各学术期刊发布GenAI使用规范的方式缺乏规范性，作者、编辑、审稿人处理稿件时不能及时获得GenAI规范的相关要求；已有规范对作者使用GenAI工具的要求较为明确，而对编辑和审稿人方面的要求较少，且各政策的具体规定内容未形成共识。 结论 社会科学领域学术期刊应从期刊、作者、编辑、审稿人4个维度制定与完善学术出版中GenAI使用规范：在期刊方面，应根据实际需要及时制定与完善GenAI使用规范、通过多渠道发布规范内容、强化对审稿人及编辑关于GenAI生成内容识别分析能力的培训、积极研制或引入AIGC检测服务并建立分级违规惩戒机制；针对作者，应明确合理使用、禁止使用、责任归属、公开披露、违规惩戒等原则；针对编辑，应明确禁止GenAI参与决策、AIGC检测、维护知识产权等原则；针对审稿人应明确保密性、独立性、透明性、监督审查等原则。

关键词: GenAI使用规范, BERTopic模型, AIGC检测, 出版伦理规范, AI伦理