Computational Normative Decision Support Structures of Forensic Interpretation in the Legal Process

A broad range of questions at various instances in the legal process can be stated and analysed in terms of formal decision theoretic models, with results conveyed in graphical terms, such as decision trees. However, the real-world decision problems encountered by the participants of a legal process, including judges, prosecutors and attorneys, present challenging features, such as multiple competing propositions, variable costs and uncertain process outcomes. This complicates decision theoretic computations and the use of diagrammatic devices such as decision trees which mainly provide static views of selected features of a given problem. Yet, the issues are inherently dynamic, and the complexity of strategic planning and assessing legal tactics-given a party's standpoint-increases even further when considerations are extended to information provided by forensic science services. This is because introducing results of forensic examinations may impact on the probability of various trial outcomes and hence crucially impact on a party's interests. In this paper, we analyse and discuss examples of decision problems at the interface of the law and forensic science using influence diagrams (i.e., Bayesian decision networks). Such models, hereafter called normative decision support structures, can be operationally implemented through commercially and academically available software systems. These normative decision support structures represent core computational models that can be integrated as part of decision and litigation support systems, to help the participants of a legal process answer a variety of questions regarding complex strategic decisions.