摘要
政府向第三方机构下放评审权,标志着医院分级管理工作从基于既定规则的等级评审向基于统计回归的专家排序转变,但单纯的患者声誉参考显然无助于重大公共卫生事件中风险评估先行的监管实践。数字抗疫的稳步推进,或将疫情防控常态化时人们对算法治理的主观想象变为现实,其合法性取决于"升格推演"的指标求解方式能在多大程度上缓解监管资源稀缺并避免道德危机。可解释性的法律要求、规范续造的边界限制,框定了机器学习模型"生成式"构架的仿生路径以及由排序算法向聚类和分类算法递变的必然趋势。公共卫生领域算法治理的实现,还需以消除行业数据共享限制的宏观政策、消解编译偏差的中观价值谱系、消弭算法歧视的微观条例规章作为法律保障。
The decentralization of inspection power from the government to third-party institutions marks an approach transition from traditional grade rating to expert judgment in prioritizing hospital monitoring, yet mere hospital reference helps little in the regulators’ risk-based approach in the public health sector. The epidemic situation of the coronavirus propels the Digital Disease Resistance Scheme which highlights a batch deployment of AI applications;justified by its promise to alleviate resource scarcity and avoid moral crisis, algorithm governance in the public health sector is no longer just a vision. Compliance requirements of interpretability and legal boundaries of Rechtsfortbildung frame the structural path of machine learning;ranking algorithm may gradually give its way to clustering or sorting algorithm in the long run. This paper concludes by proposing a three-layer legal "infrastructure"—— macro-scopic guiding principles, meso-level value hierarchy and sector-specific responsibilities — to safeguarding an algorithm governance future in the broad sense.
出处
《法学评论》
CSSCI
北大核心
2021年第3期95-107,共13页
Law Review
基金
国家社科基金青年项目“自动化应用提升现代化治理的法律保障研究”(项目号:20CFX006)阶段性成果
关键词
算法治理
人工智能
分级评审
风险评估
法律保障
Algorithm Governance
Artificial Intelligence
Grading Appraisal
Risk Assessment
Legal Safeguard