论大数据证据的周延二分——基于实物—言词证据的二分体系  

The comprehensive dichotomy of big data evidence: An analysis based on the dichotomy between physicalevidence and testimonial evidence

作　　者：邓皓元 Deng Haoyuan(Faculty of Law,University of Freiburg,Freiburg im Breisgau 79110,Germany.)

机构地区：[1]弗莱堡大学法律系,德国弗莱堡79110

出　　处：《证据科学》2024年第4期435-450,共16页Evidence Science

基　　金：国家留学基金国家建设高水平大学公派研究生项目(CSC202308080060)。

摘　　要：大数据集体量庞大,无法由人力遍历,只能经算法加工形成大数据报告后作为证据材料使用。根据算法加工的逻辑本质,亦即是否引入大数据集以外的信息,可把大数据报告周延地二分为分析性和综合性两类。此二分法与实物—言词证据二分体系相契合。实物证据的本质特征是提供直接性的感知;认知工具只要能忠实反映证据信息并在符合证明目的的范围内不造成证据信息减损,就不影响这种感知的直接性。分析性大数据报告提供直接感知,本质上属于实物证据。言词证据的本质特征是提供间接性的感知;事实认定者无法就待证事实形成亲身感知,只能借助证人、鉴定人的言说间接地形成感知。综合性大数据报告提供间接感知,本质上属于言词证据。作为实物证据,分析性大数据报告与大数据集的符合性可由事实认定者直接校验。作为言词证据,综合性大数据报告与大数据集的符合性判断无法脱离对算法的解释;由于算法的自我解释不可信赖,这一解释只能仰赖鉴定人。The massive volume of big data cannot be fully inspected by humans, but can only be convertedby algorithms to big data report, which can be used as evidence. Depending on whether the algorithm incorporatesinformation from outside the big data set, big data report can be comprehensively divided into two categories: analyticalreport and comprehensive report. This classification aligns with the dichotomy between physical evidence andtestimonial evidence. Physical evidence is characterized by providing direct perception of facts. As long as cognitivetools accurately reflect the evidential information and do not cause evidential information’s impairment within the scopeof the purpose of proof, the directness of such perception will not be affected. Analytical big data report provides directperception and thus is essentially physical evidence. On the other hand, testimonial evidence provides indirect perception of facts. The factfinder cannot directly perceive facts by their own perceptions, but can only relies on witnesses’ andexperts’ statements to form indirect perception of facts. Comprehensive big data report provides indirect perception and isessentially testimonial evidence. The conformity of analytical big data report, which is deemed as physical evidence, withthe big data set can be verified directly by the factfinder. However, the assessment of the conformity of comprehensivebig data report, which is deemed as testimonial evidence, with the big data set cannot be separated from the interpretationof the algorithm. Since the algorithm’s self-explanation is unreliable, such interpretation should be conducted by experts.

关 键 词：大数据证据 实物证据 言词证据 算法 生成式人工智能 

分 类 号：D918.9[政治法律—法学]