算法自动化决策的属性、场域与风险规制

Algorithmic Automated Decision Making of Attribute,Field and Risk Regulation

智能算法升级迭代,使得自动化决策技术广泛应用于公、私法两大领域,它在方便经济生活和社会管理的同时,也因对个人自决权应用界域的侵蚀而导致各种担忧。自动化决策算法本质上是一种数理智能程序运算,辩明算法决策的法律属性,厘清人工智能作为“人格性工具”的面相,揭示算法决策与法律行为的实质区别,有利于科学地认识算法决策的法律效果。尽管决策算法的运行逻辑结构在形式上是数据的输入和决策结果的输出,但算法逻辑是二者相关关系的判断,因此,涉及道德和伦理等价值评价的决策,人类不应该“放权”于算法。在算法决策应用中有必要界分人工决策和算法决策的合理场域,构建二者区分标准并协调和优化决策机制。应基于决策二分法,立足规制风险目标,阐述算法决策风险的产生机理,分清算法决策风险类型,从而构建“立法+监督+协同”的算法自动化决策治理体系。

Intelligent algorithm upgrades and iterates,making automatic decision technology widely used in two fields of public and private law.It not only facilitates economic life and social management,but also causes various concerns because of the erosion of the application domain of individual self-determination.In essence,automatic decision algorithm is a mathematical intelligent program operation.It is helpful to understand the legal effect of algorithmic decision scientifically by clarifying the legal attribute of algorithmic decision,clarifying artificial intelligence as a “personality tool” and revealing the substantial difference between algorithmic decision and legal action.Although the operation logic structure of decision algorithm is the input of data and the output of decision result in form,the algorithm logic is the judgment of the correlation between the two,not the causal relationship in the sense of reasoning.Therefore,human beings should not “delegate” power to algorithms for decisions involving moral and ethical evaluation.In the application of algorithmic decision making,it is necessary to divide the reasonable field of artificial decision making and algorithmic decision making,construct the distinguishing standard of the two and coordinate and optimize the decision making mechanism.Based on the dichotomy of decision and the goal of regulation risk,this paper expounds the mechanism of algorithmic decision risk and distinguishes the types of algorithmic decision risk,so as to construct the algorithmic automatic decision governance system of “legislation + supervision + coordination”.