Developing Evidentiary Foundation Based on Assessment of Forensic Results

Abstract: Issues of development of modern approaches to probabilistic assessment of forensic research results have been receiving increased amounts of attention owing to the need for clear characteristics of the limitations of research results, which include indicators of uncertainty of obtained data and associated estimated probabilities. In the modern theory of judicial evidence assessment, the use of probabilities is acceptable and even preferred, and one of the main provisions is the principle of comparing probabilities in light of their dependency on competitive versions that arise from the adversarial nature of court proceedings. In this regard, the purpose of this article is to develop methodological approaches to the use of the likelihood ratio as the most appropriate way of determining the significance of conclusions submitted to the court by forensic experts for the formation of evidentiary foundation. A brief overview of publications from 2000–2018 concerned with forensic applications of the concept of the likelihood ratio is given. This concept can be used to reliably assess the credibility of evidence. In this article the term “evidence” is considered as various continuous quantitative measurements (of properties and features of forensic objects) that are used to compare a known sample with a questioned one and establish whether they originate from the same source or from different sources. The article discusses the most common normal distribution of continuous data and a general approach to calculating the likelihood ratio (LR) using probability density functions (pdfs). It is shown that accounting for the variability of the compared samples when calculating LR requires three databases: a potential database, a control database of the known sample, and a comparative database of the questioned sample. Examples of calculating LR and the strength of evidence for various types of examinations are given. The procedures for calculating LR are generally similar, but the authors suggest different techniques of calculating and graphically representing the strength of evidence. The value of the so-called cost of (penalty for) incorrect results (ClLR), is considered in detail; the concepts of its validity and reliability, as well as its credibility interval, are introduced. The article highlights a number of specific features of calculating LR for multivariate continuous data. Of great interest is the forensic audio analysis application of speaker models in the form of weighted sums of Gaussian densities of M components (Gaussian mixture models, GMM), where each component is a D-dimensional Gaussian pdf with an average vector value and a covariance matrix. It can be assumed that the use of GMM pdfs for calculating LR is effective not only for forensic audio examination but also for other types of examinations. The universal applicability of using the likelihood ratio to assess the similarity/difference of forensic objects indicates high viability of the approach. © 2021, Pleiades Publishing, Ltd.