A Literature Review of Research on Algorithmic Evidence in China

Abstract

As big data, machine learning, and automated analytics increasingly enter the field of criminal justice, algorithmic evidence has become a frontier issue in Chinese evidence scholarship. Existing studies mainly revolve around two closely related questions. The first is whether algorithmic evidence should be accommodated within existing statutory categories of evidence, such as electronic data, appraisal opinions, and specialized-issue reports, or instead be understood as an independent new form of evidence. The second is how algorithmic evidence should be distinguished from adjacent concepts, including electronic data, big-data evidence, AI evidence, and scientific or expert evidence. A systematic review of representative Chinese scholarship shows that existing research has already revealed, with considerable clarity, the technical dependence, derivative character, professional complexity, and black-box risks of algorithmic evidence, but has not yet established a stable connection between conceptual characterization and rule construction. This article argues that neither subsuming algorithmic evidence wholesale under any single existing evidentiary category nor simply declaring it a new independent category can adequately capture its normative complexity. A more persuasive approach is to understand algorithmic evidence as a two-tier structure composed of a raw-data layer and an algorithmic-conclusion layer. This framework better explains the principal disagreements in the existing literature and helps clarify the relationship between algorithmic evidence and related concepts. On that basis, future research should move beyond disputes over labels and focus instead on the systematic design of review rules, particularly with respect to admissibility, reliability, explainability, modes of challenge, and procedural safeguards.