5.12 Serial evidence schemes
An example of a simple graphical model with multiple pieces of evidence is shown below (Koeijer et al. 2020):Figure 5.3: An example of a serial evidence scheme in a graphical model. Nodes on the graph depend upon previous nodes and this dependence should be accounted for in probability and LR calculations.
The graphical model in Figure 5.3 shows a generic example of linking together the suspect, the forensic evidence, the physical items to which the evidence relates, and a proposition of interest to the case - in this case an activity. Each type of object on the graph is represented with a different shaped box in order to clearly show that they differ in nature. For example, physical items to which the forensic evidence relates are shown within a rectangular box, whilst the proposition of interest is shown within a diamond-shaped box. The edges indicate the forensic induction process between the objects. For example, \(E_1\) above provides a link between the suspect and item 1 whilst \(E_2\) provides a link between item 1 and item 2. etc. leading to a logical sequence that links the suspect to the activity.
The graphical model in Figure 5.3 also shows a specific type of evidence scheme. Objects along this type of scheme logically depend upon those which come before them in order to construct the sequence in the graph. For example, without \(E_2\) there is no longer a chain of evidence between the suspect and the activity since there is no longer a link between item 1, which we can directly link to the suspect, and item 2. This type of evidence scheme is known as a serial evidence scheme (Koeijer et al. 2020) because of the serial nature of the evidence chain that it represents. This logical structure of the evidence must be accounted for in the probability model for this evidence chain and in the LR calculations. This graph indicates that \(E_1\) and \(E_2\) cannot be considered as statistically independent conditional on the case circumstances and propositions because there is a clear dependence between \(E_1\) and \(E_2\) through item 1. The probability calculations to model these dependencies in order to construct an LR can be complex and often use advanced methods such as Bayesian Networks. Bayesian Networks provide a probabilistic graphical model for handling multiple evidence types and an introduction to them can be found in Roberts and Aitken (2014).