自主算法隐私保护的规范与技术分析

Normative,and Technical Analysis of Privacy Protection of Autonomous Algorithms

自主算法不同于计算机化的初阶算法,具有不可控制性和不可预测性。由于其随机性工具、线上参数动态变化、远超于人的计算力及逐步走向无监督的发展趋势,我国现行民法规范,无法证成自主算法行为侵犯了用户的隐私权。这将导致一系列问题,如相同损害无法获得相同法律救济的不平等性、个人与技术之间的经济地位严重失衡以及文化伦理等“隐性价值”在算法技术中的缺失等。为解决现行隐私保护的法律标准不够灵活且不够中性的理论困境,我国应重构自主算法隐私保护的评价标准。该标准的内在要求是动态性、客观性和相当性。借鉴美国的经验,认定自主算法侵犯隐私权时,隐私应采动态组合标准,过错应采合理人标准,因果关系应采可预见性标准。

Autonomous algorithms are different from computerized primary algorithms,which are uncontrollable and unpredictable.Due to the random tools,dynamic changes of online parameters,computing power far beyond human beings,and a gradual trend toward unsupervised development,China’s current civil regulations cannot legalize autonomous behaviors that infringe on users’personal privacy rights by autonomous algorithms.This will lead to a series of problems,such as the inequality that the same damage cannot obtain the same legal remedy,the serious imbalance between the economic status of individuals and technology,and the lack of“hidden value”such as cultural ethics in algorithm technology.In order to solve the theoretical dilemma that the current legal standards for privacy protection are not flexible and neutral enough,our criteria should reconstruct the evaluation criteria for privacy protection of autonomous algorithms.The criteria should have three indicators of dynamics,objectivity and equivalence.Drawing on the experience of the United States,when determining that autonomous algorithms infringe the right to privacy,the determination of privacy should adopt the dynamic combination standard,the fault should adopt the reasonable human standard,and the causality should adopt the predictable standard.