The forensic science is undergoing a paradigm shift from the traditional paradigm to the features LR paradigm and the similarity scores LR paradigm, in an era of parallel three paradigms. Given the advantages and development opportunities brought by Bayesian LR paradigm, paradigm shift has become a major trend in forensic science in the global. However, paradigm shifts have not yet been fully realized on a global scale, and the development of the forensic paradigm is imbalance in various forensic discipline and in different countries and regions. The main reasons that hinder the paradigm shift of forensic science include limitation of the technology methods of new paradigm, misconceptions of some personnel about the new paradigms, insufficient abilities related the new paradigms, and legal application issues. Except for DNA evidence identification, the application of new paradigms and paradigm shift in China are relatively lagging behind. This article proposes an implementation path for the paradigm shift of forensic science in China, including scholars and practitioners conducting scientific research on new paradigms, regulatory authorities making paradigm shift strategies and plans, forensic lab developing and confirming LR methods, developing LR verbal scales, collecting data, training examiner and taking proficiency tests, decision-makers receiving relevant education, and legislators adjusting relevant regulations.

The likelihood ratio paradigm of facial similarity score is the theory and method for interpreting the evidential significance of score finding from facial comparison. Facial similarity score likelihood ratio is the latest method of Bayesian likelihood ratio paradigm for forensic science. The likelihood ratio (LR) of facial similarity score is the ratio of the occurring probability of the facial score finding quantitatively assigned based on the probability distribution data of facial scores, under a pair of conflicting propositions that usually represent the claims of the prosecution and defense parties. The propositions typically deal with the question of whether a facial image with unknown identity collected at a crime scene comes from a suspect with known identity. The face score LR expresses the relative support direction and strength of the face score finding for the propositions of the prosecution and defense parties, providing quantitative evidence value for decision-makers to determine the disputed fact of the face source. The decision maker determines the fact of the facial source proposition based on the facial score LR opinion, or the posterior probability of the facial source proposition derived from the LR and the prior odds through Bayesian law, combined with other evidence, to exclude reasonable doubt. The likelihood ratio paradigm of facial similarity score is completely different from the traditional paradigm we are accustomed to in terms of scientific logic, opinion formation, expression, understanding, and reasoning applications. It also differs from the widely used LR paradigm of DNA feature findings, which poses new requirements and challenges for forensic examiner and decision-makers.

The subjective likelihood ratio paradigm of pattern features is the theory and method for the interpretation of evidence significance of feature findings of pattern evidence. Subjective likelihood ratio (LR) of pattern features is the probability ratio of occurrence of pattern features assigned under the two opposing propositions representing the prosecution and defense respectively, based on expert knowledge or the combination of expert knowledge and data. The proposition hierarchy for evaluating the subjective LR of pattern features includes source level and active level. Subjective LR of pattern features expresses the relative support direction and intensity of the pattern features findings for the propositions of the prosecution and defense, providing qualitative evidence value for decision makers to determine disputed factual propositions. Decision makers will determine the propositions facts based on subjective LR opinions of pattern features, or based on the posterior probabilities of the propositions derived from LR through Bayes’ theorem, combined with other evidence in the case, in the way to exclude reasonable doubts. The subjective LR paradigm of pattern features differs significantly from the traditional paradigm we are accustomed to in terms of scientific logic and the formation, expression, understanding, and reasoning application of opinions, and it also differs from the objective LR paradigm of DNA feature. This presents new requirements and challenges for forensic examiner and decision-makers.

The objective likelihood ratio (LR) paradigm of DNA feature is the theory and method for interpreting the evidential significance of DNA analysis findings. It is a typical representative of the forensic Bayesian likelihood ratio paradigm. The DNA feature’s objective LR is the ratio of the probability of DNA feature findings quantitatively assigned based on model and data, in the context of two alternative propositions typically advocated by both of the prosecution and defense. The hierarchy of propositions includes the sub-sub source level, sub-sources level, sources level, and activity level. The LR of DNA feature findings expresses the relative support direction and strength of the DNA findings for the propositions of the prosecution and defense, providing quantitative evidence value for decision-makers to determine disputed propositional facts. Decision makers will determine the propositional facts without reasonable doubt, based on the LR opinions of DNA findings or the posterior probability of the propositions derived from LR opinions and Bayesian laws, and combined with other evidence in the case. The DNA feature’s objective LR paradigm is completely different from the traditional paradigm that we are used to in terms of scientific logic, opinion formation, expression, understanding, and reasoning applications, which poses new requirements and challenges for forensic examiner and decision-makers.

The establishment of a scientific quantitative evaluation system for fingerprint evidence, especially how to introduce the statistical method of likelihood ratio into the digital representation of fingerprint identification, is a hot issue in the current theoretical and practical research of forensic science. The construction of a scientific and effective likelihood ratio evaluation model for fingerprint evidence requires rich same-source and different-source fingerprint databases to obtain the likelihood function with a stable distribution law, thus, the quality of the same-source and different-source databases directly affects the performance of the likelihood ratio model. By using the live-scan fingerprint collector and screen recording software to obtain more than 1 000 distorted fingerprint images for each fingerprint in different distortion modes, a total of 200 000 same-source fingerprints are obtained from 200 simulated fingerprints, which constitutes the same-source fingerprint database; and the different-source fingerprint database consists of ten million people’s ten-fingerprint database in policing practice. On this basis, the automatic fingerprint identification system is utilized for query and comparison, and the comparison score data are evaluated. The experimental results show that the fingerprint data of different distortion modes have significant differences; the degree of pressure and the impressing time have little effect on the comparison scores of fingerprints. From the results of statistical analysis of the total number of samples and the subsample data after different degrees of reduction, it can be seen that the number of same-source samples of each fingerprint can still form a stable distribution law when the number of fingerprints is as few as 155. Therefore, the database we built is rich in the number of same-source and different-source fingerprints, reasonable in structure, and has the data basis for forming a stable distribution law, which is helpful for the subsequent establishment of the likelihood ratio evaluation model.

The interpretation paradigm of forensic findings is undergoing a shift from the traditional paradigm of categorical conclusions to the likelihood ratio paradigm of forensic findings. The forensic likelihood ratio paradigm, DNA evidence as a typical one, has improved the validity of forensic opinions and transformed the reasoning and proof mode of forensic opinions. The forensic likelihood ratio paradigm is based on the Bayesian likelihood ratio framework. The likelihood ratio (LR) of the forensic findings is assigned under a pair of opposing source or active propositions advocated by the prosecution and defense, and the LR is used as the evidence value of forensic findings relatively supporting the propositions of both prosecution and defense parties. It provides LR opinions of forensic findings to help decision-makers infer and determine disputed propositional facts on the source or active level. The forensic likelihood ratio paradigm is completely different from traditional paradigms in terms of scientific basis, opinion formation, expression, understanding, and reasoning application, which poses new requirements and huge challenges for forensic examiner and decision-makers in criminal proceedings.

The forensic paradigm is the scientific theories and methods used in the process of interpreting the findings of forensic examination and forming expert opinion. There is a paradigm shift from the traditional paradigm of categorical source conclusions to the paradigm of evaluative opinion. The traditional forensic paradigm is based on the assumption of feature uniqueness. The traditional paradigm of forensic science has a history of over 100 years of development and application, and has been applied to almost all physical evidence except DNA evidence. After detecting and comparing trace evidence from crime scene and known source sample, examiner will determine whether the features of the trace evidence match features of the sample, and will use threshold decision-making to give opinions on the trace evidence and the sample came from a same source or from different sources. In the traditional paradigm, the process by which examiner forms a categorical source opinions from results of the features is a deductive reasoning process: the major premise is the assumption of the uniqueness of trace features, the minor premise is the results of feature matching (or no-match), and the conclusion is that the trace and sample has same (or different) source. As long as the major and minor premises are true, the categorical opinion on source of the traditional paradigm is correct. However, with the development and maturity of evaluative methods for forensic DNA results, some scholars questioned the lack of empirical proof for the hypothesis of feature uniqueness in the traditional forensic paradigm, and thus believe that deductive reasoning without a major premise of the assumption of the trace features uniqueness has no validity, and therefore, the categorical source opinion in the traditional paradigm lacks a solid scientific foundation.

Fingerprint, one kind of the most important forensic evidence, is capable of having an individual identified. Therefore, it has played a crucial role in a police investigations and court litigation since it was admitted under jurisprudence. Regarding the verdict of fingerprint identification, there are currently only three propositions in China: recognition, exclusion, and inconclusiveness. Presumably, such handling roots its basis on the experience and practical situation of China's crime prevention and court processing hitherto, yet having caused abandonment or unusedness of amount-huge fingerprints collected of less than 8 minutiae from crime scenes due to their disqualification to the requirements of source threshold for fingerprint identification. However, these fingerprints are significant for police investigation and court processing. Thus, a probabilistic approach was described here with mathematic modeling to count minutiae by the related fingerprint image divided into fan-shaped sectors. Based on the statistics of 15 million fingerprint images, a function of probability density was fitted into fingerprint minutiae of all fan-shaped sectors and then modified under Bayesian Information Criteria, plus the addition of noises. Consequently, a probability was acquired towards identity recognition about a fingerprint under scrutiny. Through a trial of several examples, the results showed that the matching probability of fingerprint pairs was positively correlated to the quantity and stability of the analyzed minutiae yet negatively to the incidence of minutiae occurring inside the fan-shaped sectors. This study provided a novel attempt to rediscover the evidential value of those ‘useless fingerprints' displaying no sufficient details for identification but frequently found at crime scenes. Such an approach should be a crucial step for fingerprint identification from quality to quantity analysis, having significant potential to identify a criminal in combination with the quantifying applications of other forensic evidence.