马明亮 王士博

（中国人民公安大学，北京 100038）

【摘 要】进入数字时代，在视听资料、电子数据的证据种类中已然分叉出新的证据类型—大数据证据。基于不同的底层技术支撑，大数据证据与传统意义的电子数据、鉴定意见有着本质区别。但遗憾的是，学界与实务部门并没有充分意识到这一潜在变化及其可能带来的证据制度变革。实践中，司法人员对大数据证据的审查判断保守地依附于传统证据类型的审查规则或印证结论。这种依附性证明力规则会带来诸多弊端：在事实查明方面，大数据证据真实性的认定科学性与合理性不足；在诉讼结构方面，因缺乏实质意义审查而径直采信大数据证据，导致侦查中心主义再次抬头。鉴此，立法需要根据大数据证据的本质特征确立独立的证明力规则，作为自由心证的外在界限。根据算法作用方式的不同，大数据证据分为基于海量数据库比对生成的大数据证据和基于算法模型的大数据证据两种类型。对于前者的证明力审查，考虑到其所具有的概率性，应当建立基于贝叶斯定理的似然率评价规则，并围绕算法是否具有稳健性的判断标准建立补强规则。对于后者则应当建立行为数据规律“前理解”的审查规则，并依据海量数据的来源对证明力强度予以区分。如此，形成大数据证据证明力的独立审查规则，使大数据证据在数字时代的真实性认定具有了更加科学的基础，进而确保实质意义的审判中心主义。

【关键词】大数据证据；证明力规则；依附性审查；似然率；算法模型

【中图分类号】D915.13

【文献标识码】A

【文章编号】1674-1226（2021）06-0645-12

On the rules of probative value of big data evidence. Ma Mingliang, Wang Shibo, People's Public Security University of China, Beijing 100038, China.

【Abstract】In the digital age, big data evidence, as a new type of evidence, has emerged among the types of audiovisual materials and electronic evidence. Based on different underlying technologies, big data evidence is essentially different from traditional electronic evidence and appraisal opinions. However, it is a pity that the academic and practical circles are not fully aware of this potential change and the change of evidence system it may bring about. In practice,the judicial personnel’s review and judgment of big data evidence conservatively depends on the reviewing rules of traditional evidence types or corroborating conclusions. The rules of probative value with dependent nature will lead to many disadvantages: in the aspect of fact finding, the confirmation of the authenticity of big data evidence is not scientific and reasonable enough; in the aspect of lawsuit structure, big data evidence was directly accepted due to lack of substantive review , resulting in a resurgence of investigation centrism. In view of this, independent rules of probative value according to the essential characteristics of big data evidence should be established, as the external boundary of free evaluation of evidence. According to the different action modes of algorithms, big data evidence can be divided into two types: big data evidence based on massive database comparison and big data evidence based on algorithm model. For the probative force review of the former, considering its probabilistic characteristics, the likelihood ratio evaluation rules based on the Bayes Theorem should be established, as well as the evidence rules of corroboration based on the criterion of whether the algorithm is robust. For the latter, the review rules of " Pre-understanding" of behavioral data pattern should be established, and the strength of probative force should be evaluated according to the sources of massive data.In this way, the independent review rules for the probative force of big data evidence are formed, which provides a more scientific basis for the authenticity identification of big data evidence in the digital age, so as to ensure the substantive trial centrism.

【Key Words】Big data evidence; Rules of probative value; Review with dependent nature; Likelihood ratio; Algorithm model