算法辅助量刑的风险成因与规范进路——以美国实践争议为切入点

Risk Causes and Normative Approach of Algorithm-Assisted Sentencing--The Controversial US Practice as Perspective

在美国算法辅助量刑的司法实践中,就算法的辅助地位、法官自由裁量权与被告人诉讼地位等问题,存在争议。算法固有偏见及其程序不透明等内在风险,以及算法辅助量刑规范体系缺位的制度性风险,不仅严重限制了被告人的算法知情权,而且使经由算法辅助的量刑裁量权不受约束,有碍程序公正的实现。为从实体与程序上实现公正量刑,应对辅助量刑的算法进行规范化。在算法层面,应当坚持算法公开原则,明确商业秘密不得阻碍算法披露,并以一致性标准与算法留痕促进算法公平。在诉讼层面,应当对算法知情权构建具体的权利束,在算法证据规则上区分由检察机关提出与被告人异议而分配不同的证明标准与证明责任。

In American judicial practice of algorithm-assisted sentencing,the auxiliary status of algorithms,the discretionary right of judges and the litigation position of defendants is controversial.There are potential risks associated with algorithmic tools,including intrinsic risks such as the inherent bias of algorithms and the opacity of their procedures,as well as the institutional risk of the absence of a normative system for algorithm-assisted sentencing.These risks not only severely limit the defendant's right to be informed of the algorithm,but also leave the sentencing discretion assisted by the algorithm unrestrained and hinder the achievement of procedural justice.In order to achieve fair sentencing both in substance and in procedure,the algorithm-assisted sentencing should be regulated.At the algorithmic level,the principle of algorithmic openness should be adhered to,and it is supposed be clear that trade secrets should not prevent the algorithm from being disclosed,and the algorithmic fairness should be promoted with standards of consistency and algorithmic traces.At the level of litigation,a specific bundle of rights is required to be constructed for the right of algorithmic information,and the rules of evidence for algorithms distinguish between the standard of proof and the burden of proof allocated by the prosecutor that raises the evidence and the defendant that objects.