数字时代风险分级视域下的算法透明治理  

Algorithm transparent governance from the perspective of risk grading in the digital age

作　　者：陈文兴[1] 张羽芊 CHEN Wenxing;ZHANG Yuqian(College of Public Administration and Law,Fujian Agriculture and Forestry University,Fuzhou 350002,China)

机构地区：[1]福建农林大学公共管理与法学院,福建福州350002

出　　处：《南京邮电大学学报(社会科学版)》2025年第2期42-49,共8页Journal of Nanjing University of Posts and Telecommunications(Social Science Edition)

基　　金：福建省社科基金项目(FJ2021BF015)。

摘　　要：算法黑箱、算法歧视等负外部性引发了对算法透明的迫切需求,然而算法技术往往涉及算法持有者的商业秘密,不同的利益取向诱发了算法透明与商业秘密保护之间的紧张关系。风险分级既契合算法透明对于算法安全及算法可信的治理要求,又满足商业秘密保护的发展需求,是平衡安全与发展的重要之策。目前算法透明与算法分级均止步于原则性要求,对算法的未知恐惧催生了对算法透明的盲目追求,因此有必要摒弃全有或全无的路径依赖,在风险分级治理视域下搭建算法有限透明之架构,结合风险导向的源头进行治理。The negative externalities such as algorithm black box and algorithm discrimination have raised an urgent need for algorithm transparency.However,algorithm technology often involves the trade secrets of algorithm holders,and different interest orientations have induced a tense relationship between algorithm transparency and trade secret protection.Risk classification not only meets the governance requirements of algorithm transparency for algorithm security and trustworthiness,but also satisfies the development needs of trade secret protection,making it an important strategy to balance security and development.At present,both algorithm transparency and algorithm grading are limited to principle requirements,and the fear of the unknown in algorithms has led to blind pursuit of algorithm transparency.it is necessary to abandon the path dependence of all or nothing,and develop a limited transparency architecture for algorithms in the perspective of risk grading governance,combined with risk oriented source governance.

关 键 词：算法透明 商业秘密保护 风险分级 自动化决策 监管沙盒 算法治理 

分 类 号：D922.1[政治法律—宪法学与行政法学]