论产品责任在人工智能辅助医疗诊断系统侵权责任认定中的应用

On the Application of Product Liability in Determining Tort Liability for AI-Assisted Medical Diagnostic Systems

摘要: 人工智能辅助医疗诊断系统的广泛应用对现行侵权责任规则提出了挑战。由于“算法黑箱”的存在，医务人员的审查能力被实质性削弱，继续沿用传统医疗损害责任有违公平，而产品责任的适用亦面临缺陷认定难与因果关系断裂的困境。本文立足于风险社会理论，论证了对此类系统适用产品责任的法理正当性。在规则构建上，建议重构产品缺陷认定标准，引入事实与法律结合的双重因果关系判定规则。最后，基于人机协作的不同模式，将侵权形态类型化为“隐秘型缺陷”、“常识型缺陷”及“人机耦合型缺陷”，并据此构建差异化的责任分担机制，以期在救济受害人与鼓励技术创新之间寻求平衡。

Abstract: The widespread deployment of AI-assisted medical diagnostic systems poses challenges to existing tort liability rules. Due to the existence of the “algorithmic black box”, the reviewing capacity of medical personnel is substantially weakened, making it inequitable to continue applying traditional rules of medical malpractice liability. Moreover, the application of product liability also faces difficulties in defect identification and the breakdown of causation. Drawing upon risk society theory, this article demonstrates the jurisprudential legitimacy of applying product liability to such systems. With respect to rule construction, it proposes reconstituting the standard for determining product defects and introducing a dual causation test that integrates factual and legal causation. Finally, based on different modes of human-machine collaboration, it typologizes infringing acts into “latent defects”, “common-sense defects”, and “human-machine coupling defects”, and accordingly constructs differentiated liability allocation mechanisms, with a view to striking a balance between victim compensation and the encouragement of technological innovation.