Reliability Risks and Regulation of Algorithmic Evidence in Criminal Certification

Algorithmic evidence is the product of the deep integration of algorithm and criminal justice. In criminal proof, the reliability risk of algorithmic evidence mainly focuses on three levels: at the level of legitimacy, the definition of evidence qualification is vague, the data collection is illegal, and the algorithmic technology application program is not standardized; at the level of authenticity, the data source is biased, the algorithm is discriminatory, and the black box dilemma; and at the level of relevance, the causality is insufficient, and the predictive conclusions are easy to be mechanically applied. In this regard, the reliability of should be guaranteed through the legalization of evidence types and the traceability of data sources; the establishment of algorithmic technology visualization system and third-party verification mechanism to achieve technical reliability; and the clear logical connection between the conclusion and the facts to be proved and the setting up of manual review nodes to strengthen the logical reliability. It provides institutional and technical support for the innovation of judicial proof system in the digital era.