摘要
以算法为监管手段的证券数字监管,可以通过对证券业务数据样本的处理、分析,找出异常交易行为下隐藏的不法意图,为证券监管机构提供事前、事中、事后监管措施。但监管规则含义泛化、数据质量标准不一、算法非中立性和“黑箱”的存在,给算法带来了法律风险,影响了算法推理、决策和结果在法律程序中的地位和作用。从英国金融行为监管局(FCA)的实践经验来看,设计机器可执行规则是提高证券监管效率的重要手段;统一数据标准是提升算法可信度的重要措施;算法合法化监管是保证算法中立的关键。结合我国证券违法活动的算法监管现状,可从三个层面完善算法嵌入监管的法治化路径,一是在规则层面,实现算法代码与监管规则的深度融合;二是在数据治理层面,完善算法输入层和识别层的数据治理程序;三是在法律程序层面,建立算法审计机制和算法可解释性机制。
For digital scheme of securities supervision with algorithm as the regulatory means,the hidden illegal intentions under abnormal trading behaviors could be found out through the processing and analysis of securities business data samples,to provide the securities regulators with advanced,operational and post regulatory measures.Due to the generalization of the meaning of regulatory rules,inconsistent data quality standards,nonneutrality of algorithms and the"black box",legal risks are introduced to algorithms.This situation also affects the status and role of algorithmic reasoning,decision-making and results in legal procedures.From the practical experience of FCA,designing machine executable rules is significant to improve the efficiency of algorithm supervision.Data standard unification is an important measure to improve algorithm reliability.Regulation of algorithm legalization is the key to ensure algorithm neutrality.Considering the current situation of algorithmic supervision in response to illegal securities activities in China,the legalization path of applying algorithms to supervision can be improved from three levels.Firstly,at the rule level,the deep integration of algorithm code and regulatory rules can be realized;Secondly,at the level of data governance,the data governance procedures of algorithm input layer and identification layer can be improved;Thirdly,at the level of legal procedures,the algorithm audit and interpretability mechanism can be established.
出处
《金融监管研究》
北大核心
2023年第11期94-113,共20页
Financial Regulation Research
基金
对外经济贸易大学研究生科研创新基金资助项目“证券监管领域中算法证据的证明机理问题及程序性风险规制研究”的资助,项目编号:202372
关键词
算法监管
算法画像
数据标准
算法审计
Algorithm Supervision
Algorithm Portrait
Data Standard
Algorithmic Audit
