ABSTRACT
This chapter focuses on DNA evidence and provides examples of the use of object-oriented Bayesian network (BN) to handle cases of criminal identification and simple and complex disputed paternity. It shows how the methods can be extended to deal with more complex cases, where a crime trace may contain a mixture of DNA from more than one contributor, in varying proportions. The chapter explores some of the simplifying assumptions made so far, to account for such realistic complications as uncertainty about allele frequencies and heterogeneity in the reference population. It describes by means of a fictitious example the way in which different elements in a case can be drawn together into a single coherent story structured as a BN. The chapter argues that how a BN can be used to simplify the specification and manipulation of probabilities, in particular the use of “evidence propagation” to compute conditional probabilities taking the evidence into account.
