Bayesian Networks for Probabilistic Inference and Decision Analysis in Forensic Science, 2nd Edition

Chapter 8Aspects of combining evidence

8.1 Introduction

In the previous chapters, Bayesian networks have been constructed and discussed with the aim of approaching uncertainties that affect inference based on results of various kinds of scientific examinations, conducted on traces, such as fibres and glass, or biological material (DNA). In this chapter, such results are more generally termed evidence, or items of evidence, because the focus relies on the joint assessment of multiple results pertaining to trace items of, possibly, different natures. This slight shift in wording (from ‘findings’ to ‘evidence’) with respect to previous chapters is justified by the fact that the joint assessment of several scientific results is typically operated at an advanced stage in a judicial process (e.g. in Court), where more formal decisions are made as to which scientific results are regarded as ‘evidence’.

The networks developed so far in this book provide valuable assistance in addressing some of a wide range of issues that affect a coherent evaluation of probative value. It was emphasized that existing probabilistic solutions proposed in the scientific literature may be used as a guide to elicit appropriate network structures. By providing an explicit representation of the relevant variables together with their assumed dependence and independence properties, Bayesian networks have the potential to clarify the rationale behind a given probabilistic approach, in particular, formulae for ...