关键词: DNA证据, 证据评价, 似然比

Abstract: By combination of quantitative evidence support and the five-level propositional framework, the Bayesian Likelihood Ratio model is considered to be the mainstream international scientific evidence evaluation paradigm. In China, the application of Bayesian Likelihood Ratio model in the evaluation of DNA evidence is facing a triple challenge: At the technical level, there are problems such as difficulties in the separation of mixed samples, allele dropout, and model uncertainty; At the institutional level, the current normative documents regard uncertainty as a technical flaw, forcing appraisers to issue “pseudo-certainty” conclusions; At the cognitive level, the fact finder is prone to fall into “numerical absolutes” due to the lack of statistical knowledge; At the institutional level, the current regulatory documents regard uncertainty opinions as technical defects, forcing experts to issue “pseudo-certainty” conclusions; At the cognitive level, the fact-finder is prone to fall into “numerical absolutism” due to lack of statistical knowledge. Based on this, the study proposes a multi-dimensional improvement path: developing high-precision algorithms and optimizing mixed-sample analysis, constructing dynamic genetic databases to reduce model bias, revising judicial norms to recognize the reasonableness of uncertainty expression, and establishing a hierarchical review standard for “technical uncertainty” and “cognitive uncertainty”. The aim is to improve the practical application value of DNA evidence and promote the accuracy and fairness of scientific evidence evaluation in court.

Key words: DNA evidence, evidence evaluation, likelihood ratio