摘要：

目的 系统揭示中国科技期刊卓越行动计划（以下简称“卓越行动计划”）二期入选英文期刊与中文期刊在人工智能（artificial intelligence， AI）使用政策方面的差异特征与深层成因，为推动国内科技期刊AI治理能力规范化提供实证依据。方法 以390种卓越行动计划二期入选期刊（英文200种、中文190种）为研究对象，构建“政策来源四分类”分析框架与“政策完备度十维度”评分体系，通过访问期刊官方网站，系统采集AI政策文本，采用二值编码与描述统计相结合的方法，对中英文期刊的AI政策覆盖率、政策来源结构及内容完备度进行比较分析。结果 英文期刊AI政策覆盖率（56.5%）显著高于中文期刊（22.6%）（χ2=50.7，P<0.001），但75.9%源于出版商统一模板；在剥离被动政策后，英文期刊编辑部主动制定率仅为9.0%，低于中文期刊的22.6%。在政策完备度方面，英文主动型期刊均分4.44分，中文均分3.12分，差距集中于“署名禁止”和“图像版权”；但中文最高分（10分），反超英文分值（9分），且在“违规后果”维度领先；双方在“动态更新”维度均表现薄弱。在学科维度上，中文覆盖率差异悬殊（医学47.9%，工程技术9.0%），英文因出版商统一供给而显著收窄。结论 中英文期刊AI政策差异的根源在于出版生态的制度性约束，而非编辑部治理能力不足。英文高覆盖率源于出版商统一供给政策的制度支持，中文的学科分化反映行业组织化程度对政策扩散的决定性影响。本研究构建的评价体系经信效度与敏感性检验，具有可接受的科学性与稳健性，可为期刊AI政策评价提供可推广的测量工具。

关键词: 中国科技期刊卓越行动计划, 科技期刊, 人工智能使用政策, 人工智能政策评价, 出版生态

Abstract:

Purposes This study aims to systematically reveal the differences in artificial intelligence （AI） usage policies between English and Chinese journals included in phase II of the Excellence Action Plan for China’s STM Journals， and to explore the deep-rooted causes of these differences， thereby providing empirical evidence for improving AI governance capacity in domestic science and technology journals. Methods A total of 390 journals （200 English-language and 190 Chinese-language） selected under phase II of the Excellence Action Plan for China’s STM Journals were included as research subjects. A four-category policy source framework was constructed to classify journal AI policies into： no-policy， passive （using publisher templates）， COPE-referenced， and independently developed. Additionally， a ten-dimension policy completeness scoring system was built. AI policy texts were systematically collected by accessing the official websites of all journals， and binary coding combined with descriptive statistics was applied to compare the policy coverage rate， source structure， and content completeness between English-language and Chinese-language journals. Findings The coverage rate of AI policies in English-language journals （56.5%） was significantly higher than that in Chinese-language journals （22.6%） （χ2=50.7， P<0.001）. However， 75.9% of the English policies originated from publisher?provided templates. After excluding these passive policies， the active formulation rate by editorial offices of English journals dropped to 9.0%， which was lower than the 22.6% observed for Chinese journals. In terms of policy comprehensiveness， English active-policy journals scored an average of 4.44， while Chinese journals averaged 3.12； the gap was mainly concentrated in the dimensions of “authorship prohibition” and “image copyright”. Nonetheless， the highest score among Chinese journals （10） surpassed that of English journals （9）， and Chinese journals led in the “consequences of violations” dimension. Both groups performed weakly in the “dynamic updating” dimension. At the disciplinary level， the coverage disparity among Chinese journals was substantial （medicine 47.9%，engineering and technology 9.0%）， whereas the gap was significantly narrowed for English journals due to the uniform provision by publishers. Conclusions The root cause of the AI policy differences between Chinese and English scientific journals lies in institutional constraints within the publishing ecosystem， rather than deficiencies in editorial office governance capabilities. The high coverage rate in English journals results from institutional support through publisher-provided uniform policies， while the disciplinary divergence observed in Chinese journals reflects the decisive influence of the degree of industry organisation on policy diffusion. The evaluation framework constructed in this study， having passed tests of reliability， validity， and sensitivity， demonstrates acceptable scientific rigour and robustness， and can serve as a generalisable measurement tool for assessing AI policies in scientific journals.

Key words: Excellence Action Plan for China’s STM Journals, Scientific journals, Artificial intelligence usage policy, Artificial intelligence policy evaluation, Publishing ecosystem