This event examined the use and development of graphical modelling approaches for decision making in statistics and law. While progress has been made in the development of graphical modelling approaches for decision making in statistics and the law in the research community (e.g. use of Bayesian networks for the evaluation of forensic evidence types such as DNA, glass evidence, fibres and fingerprints), the use of these methods in practice has been limited.
Three speakers presented their perspective of the use of graphical models in statistics and the law. Professor Jim Smith presented a research perspective, speaking on the development of different graphical methods for representing competing propositions from the prosecution and the defence in criminal cases, with particular examples using Chain Event Graphs. Professor Bas Kokshoorn gave the view of a forensic scientist and highlighted how Bayesian Networks are used in practice in criminal cases in the Netherlands and what some of the challenges in doing this are. Finally, Professor Angela Daly spoke from a legal perspective, discussing some of the key legal issues that surround the use of graphical networks and other statistical approaches in real cases, for example the need for model transparency as well as training to improve understanding.
The talks were followed by a panel discussion in which questions focused on the challenges of getting legal professionals (lawyers, judges, juries) to engage with statistics, issues with ensuring the transparency of graphical modelling approaches and challenges in validating modelling approaches when acting as an expert witness with statistical evidence.