Abstract:

Purposes This paper explores the systemic impact of generative AI （GenAI） on the research ethics framework. It moves beyond the traditional “morality” perspective， which focuses on individual academic misconduct， to an “ethics” perspective characterized by collectivity. It analyzes the profound challenges arising from AI's reshaping of the research interaction environment. Methods The paper employs a methodology combining conceptual analysis and case studies. It first deconstructs the three core assumptions underpinning traditional research ethics： the singularity of the violator， the attributability of misconduct intention， and the stability of ethical norms. Second， by analyzing cases such as “prompt injection” and “writing style homogenization”， it demonstrates the profound shifts in the research ethics ecosystem caused by GenAI. Findings GenAI is systemically challenging the three core assumptions of the traditional ethical framework： First， ethical subjects are shifting from “rule followers” to “strategic actors”. Second， ethical risks are expanding from “individual intentions” to “unattributable emergent consequences”. Third， ethical norms are evolving from “static rules” to a “dynamic field”. The governance paradigm centered on preventing individual misconduct is gradually becoming ineffective， and the academic community is entering a phase of social experimentation to seek a new ethical compact. Conclusions The academic community should abandon a “containment” mindset and construct a new ethical compact within this dynamic social experiment. As key nodes in the research ecosystem， scientific journals must shift their governance strategies from preventing individual misconduct to nurturing the systemic ecosystem. It is suggested that the journal community explore three paths： establishing verifiable procedural rules to rebuild trust； moving beyond single-paper compliance checks to safeguard the diversity of the collective knowledge ecosystem； and building an evolving governance framework to lead the community in adapting to the uncertain technological environment.

Key words: Generative AI, Research ethics, Research integrity, Publication ethics, AI for Science, Knowledge production, AI+

摘要：

目的 探讨生成式人工智能（GenAI）对科研伦理框架的系统性冲击，超越传统的聚焦个体学术不端的“道德”视角，转而从集体性的“伦理”视角，分析AI重塑科研环境引发的深层挑战。 方法 采用理论思辨与案例分析相结合的方法，分析传统科研伦理框架所依赖的违规主体的单一性、不端意图的可归因性和伦理规范的稳定性3个内在假设；通过分析指令注入和写作风格同质化等案例，论证GenAI如何导致科研伦理生态的变化。 结果 GenAI正系统性挑战传统伦理框架的3大假设：伦理主体从规则遵循者转变为策略行动者；伦理风险从个体意图扩展至不可归因的涌现性后果；伦理规范从静态准则演变为动态场域。以防范个体科研不端为核心的治理范式已逐渐失效，学术共同体正进入探寻新伦理契约的社会实验阶段。 结论 学术共同体应摆脱“封堵”思维，在动态的社会实验中构建新的伦理契约。科技期刊的治理策略必须从防范个体不端转向培育系统生态。建议从3个方面进行探索：构建可验证的程序规则以重建信任；超越单篇论文的合规审查，守护集体知识生态多样性；建立动态演进的治理框架，引领学术共同体适应不确定的技术环境。

关键词: 生成式人工智能, 科研伦理, 科研诚信, 出版伦理, AI for Science, 知识生产, 人工智能+