刑事大数据证据的规范定位与审查转向——以算法不可解释性为切入  

The Normative Position and Review Shift of Criminal Big Data Evidence——Taking Algorithmic Unexplainability as an Entry Point

作　　者：林树建 Lin Shujian(School of Criminal Justice China University of Political Science and Law,Beijing 100088,China)

机构地区：[1]中国政法大学刑事司法学院,北京100088

出　　处：《北京化工大学学报(社会科学版)》2025年第1期30-39,99,共11页Journal of Beijing University of Chemical Technology(Social Sciences Edition)

摘　　要：大数据证据应用于刑事司法实践已是大势所趋。大数据证据呈现大数据集→算法→大数据报告的三元内部构造,人工智能的应用使海量非结构化数据的分析成为可能,但也伴随着算法决策不可解释的技术特征。不可解释性使大数据证据区别于言词证据,大数据集和算法的双重客观性决定了其实物证据属性。不可解释性亦造成了大数据证据的质证虚化困境,审查重心由真实性转向可靠性则能破解这一困境,具体方法是借助人类熟知的因果关系重构以概率推理为底色的算法决策。不可解释性应被视为技术特征而非技术缺陷,大数据证据的规范应对将使刑事诉讼更好地借力于科技发展,也带来了更新诉讼认识理念的契机。The integration of big data evidence is emerging as a significant trend in the future of criminal justice practice.This types of evidence is characterized by a three-part internal structure:“big dataset→algorithm→big data report”.The use of artificial intelligence enables the analysis of massive unstructured data,and it also introduces the technical challenge of algorithmic decisionmaking unexplainability.This unexplainability set big data evidence apart from testimonial evidence,while the dual objectivity of the bigdata set and algorithm establishes its property as physical evidence.Nevertheless,it also creates a challenge for the verification of big data evidence.To address this challenge,the focus of examination should shift from authenticity to reliability,with an specific approach of reconstructing algorithmic decisions—originally grounded in probabilistic reasoning—through the application of well-recognized causal relationships.Explainability should be regarded as a technical attribute rather than a flaw.Standardizing big data evidence will enable criminal litigation to better leverage technological development and provide an opportunity to update conceptual understanding of litigation.

关 键 词：大数据证据 算法 不可解释性 实物证据 因果关系 

分 类 号：D925.2[政治法律—诉讼法学]