摘要：

目的 探讨生成式人工智能（generative artificial intelligence，GenAI）介入学术出版领域引发负面连锁效应的过程，据此探索基于学术出版场景的GenAI规制进路。 方法 基于风险链条梳理负面连锁效应的形成过程，从商业利益视角分析AI商业化激化负面连锁反应的方式，再通过匹配风险特征和政策法规提出规制进路。 结果 GenAI通过11个循序渐进的阶段形成了劣币驱逐良币、认知错误循环、学术失信泛滥、学术不公放大4种负面连锁效应，AI商业化则通过固化学术成果生产方向、重塑学术研究评价体系、提高学术出版经济壁垒激化负面连锁反应。 结论 GenAI规制进路需同时关注场景适配、政策衔接与动态调整：首先，要立足学科场景、出版活动角色进行风险等级分类与责任分配；其次，要在数据治理、知识产权领域实现法律规定与实施细则的有效衔接；最后，需构建基于沙盒测试与触发式更新的动态治理机制。

关键词: 生成式人工智能, 学术出版, 负面连锁效应, AI商业化, 规制进路

Abstract:

Purposes This paper discusses the process of negative chain effects caused by generative artificial intelligence （GenAI） in academic publishing， and explores the regulatory approach of GenAI based on the academic publishing scenario. Methods Based on the risk chain， this paper sorts out the formation process of negative chain effects， analyzes how the commercialization of AI intensifies negative chain reactions from the perspective of commercial interests， and then proposes regulatory approaches by matching risk characteristics with policies and regulations. Findings GenAI has formed four negative chain effects through 11 gradual stages： bad money driving out good money， a cycle of cognitive errors， rampant academic dishonesty， and amplified academic injustice. The commercialization of AI has intensified the negative chain reactions by solidifying the direction of academic production， reshaping the academic research evaluation system， and raising the economic barriers to academic publishing. Conclusions The regulatory approaches of GenAI need to pay attention to scenario adaptation， policy coordination and dynamic adjustment at the same time： first， risk level classification and responsibility allocation should be based on subject scenarios and publishing activity roles； second， effective connection between legal provisions and implementation details should be achieved in the fields of data governance and intellectual property rights； finally， a dynamic governance mechanism based on sandbox testing and triggered updates should be established.

Key words: Generative artificial intelligence, Academic publishing, Negative chain effects, AI commercialization, Regulatory approaches