摘要：

目的 基于国际主要出版机构及学术协会公开发布的人工智能出版伦理政策文本，分析当前学术出版领域人工智能伦理治理的制度表述特征、责任配置与价值取向，为我国学术期刊完善人工智能伦理规范提供参考。 方法 采用分层目的性抽样，选取22家具有代表性的国际出版机构及学术协会公开发布的AI学术出版伦理政策为研究样本，构建“政策工具-责任主体-伦理原则”三维框架，运用定向内容分析法对政策条款的要素分布与结构特征进行量化统计与质性分析。 结果 国际出版机构人工智能伦理治理政策总体呈现较强的风险防控导向：政策工具以强制型和混合型为主；责任配置主要集中于作者端；伦理原则侧重透明度、问责、隐私保护与安全控制等程序性与合规要求；不同类型机构在规则表达和规范重点上存在明显差异。 结论 国际出版机构人工智能伦理政策初步形成以责任界定、使用披露和风险控制为重点的基本框架，在价值整合、多主体协同和制度细化层面仍存在完善空间。我国学术期刊人工智能伦理治理可从加强行业规则供给、完善多主体责任链条和推动学术诚信实质审查等方面进行优化。

关键词: 生成式人工智能, 学术出版, 出版伦理, 公开政策文本, 内容分析

Abstract:

Purposes Based on publicly available policy documents on AI publishing ethics issued by major international publishers and academic associations， this study analyzes the institutional expression， responsibility allocation， and value orientation of current AI ethics governance in scholarly publishing， with a view to providing references for improving AI ethical norms in Chinese academic journals. Methods Using stratified purposive sampling， this study selected AI ethics policies for scholarly publishing publicly issued by 22 representative international publishers and academic associations as the research sample. A three-dimensional framework of “policy instruments-responsible actors-ethical principles” was constructed， and directed content analysis was employed to conduct quantitative and qualitative analyses of the distribution of elements and structural characteristics of policy provisions. Findings Overall， AI ethics governance policies of international publishing institutions show a strong risk-prevention orientation. Policy instruments are mainly mandatory and hybrid in nature； responsibility is primarily assigned to authors； ethical principles emphasize procedural and compliance requirements such as transparency， accountability， privacy protection， and safety control； and significant differences exist among different types of institutions in rule expression and regulatory focus. Conclusions International AI ethics policies in scholarly publishing have preliminarily formed a basic framework centered on responsibility delineation， use disclosure， and risk control， but there remains room for improvement in value integration， multi-actor coordination， and institutional refinement. At present， AI ethics governance in Chinese academic journals can be strengthened by enhancing the supply of industry rules， improving multi-actor chains of responsibility， and promoting substantive review of academic integrity.

Key words: Generative artificial intelligence, Scholarly publishing, Publishing ethics, Public policy texts, Content analysis