Research on the Admissibility and Examination Rules of Algorithmic Evidence in Criminal Cases

Abstract

In the digital age, algorithmic evidence generated by the intervention of algorithmic technology in criminal justice is structurally impacting the traditional evidence rules system. Algorithmic evidence includes two basic forms: "matching" and "inductive." Inductive evidence, with its dual characteristics of generative dynamic fact-finding and machine-decision-based data processing, differs from static electronic data and expert opinions based on natural persons, and should be granted independent evidentiary status. At the admissibility level, algorithmic evidence meets the substantive requirements of Article 50 of the Criminal Procedure Law, which states that "materials that can be used to prove the facts of a case are evidence." Its admissibility must be determined through a triple test of relevance, reliability, and legality. At the review rules level, a two-tiered system should be constructed, distinguishing between admissibility review and probative value assessment: the former focuses on relevance, reliability, and legality as core dimensions; the latter can introduce mechanisms such as likelihood evaluation, corroboration rules, and expert assistance. Simultaneously, the obligation to disclose algorithmic technology should be improved, and the expert assistant system optimized to ensure the substantive nature of cross-examination in court, thereby achieving an organic unity between technological empowerment and procedural justice.