自主学习型算法共谋的事前预防与监管  

Proactive Regulation and Supervision on Conspiracies Based on Self-learning Algorithm

作　　者：王健[1,2] 吴宗泽[3] Wang Jian;Wu Zongze

机构地区：[1]浙江理工大学法政学院,浙江杭州310018 [2]浙江理工大学竞争法律与政策研究中心,浙江杭州310018 [3]美国加州大学伯克利分校法学院,美国加利福尼亚州94704

出　　处：《深圳社会科学》2020年第2期147-158,160,共13页Social Sciences in Shenzhen

基　　金：国家社会科学基金重点项目“反垄断法制裁的现代化研究”(18AFX019)的阶段性成果。

摘　　要：近年基于人工智能技术的自主学习型算法在信息推荐、动态定价情境下显现出如价格歧视、性别歧视等异常,逐渐引发学界对自主学习型算法问题的担忧。囿于自主学习型算法的不透明性及行为隐蔽性,往往难以在反垄断执法、司法层面对其予以及时、有效规制。因而须依据自主学习型算法自身特点构筑起合法、有效的事前管控路径,从技术监管、数据控制、算法审核等层面对自主学习型算法加以规制,及时遏止自主学习型算法共谋不当影响的出现和扩张。Informationized and data-based transaction pattern now enables enterprises to make business decisions efficiently through various computer algorithms on the basis of technologies such as big data and cloudcomputing.However,some of the abnormal behaviors of self-learning algorithm,like pricediscrimination or gender-discrimination appearing during information recommendation and dynamic pricing process,that have emerged in recent years also lead to growing concerns about conspiracies based on selflearning algorithm.Due to the opacity and concealment of self-learning algorithms,it's often difficult to regulate them timely and effectively at the antitrust enforcement level or the judicial level.Therefore,it is quite necessary to build a proactive regulation and supervision path according to the characteristics of the self-learning algorithm,and then regulate self-learning algorithms from multiple aspects,such as technical supervision,data control,algorithm review,etc.Only through such realignments,adverse effect caused by conspiracies based on self-learning algorithm may be timely contained.

关 键 词：人工智能 算法共谋 技术监管 事前管控 反垄断法 

分 类 号：D912.29[政治法律—民商法学] TP18[政治法律—经济法学]