A Bayesian network approach to evaluating footwear evidence.

The interpretation, evaluation and communication of forensic footwear findings has had an impact on its use and the perception of its value, although this is also an endemic issue in the wider forensic science discipline. Tools for presenting forensic evidential findings in a clear, robust and transparent manner, such as Bayesian Networks, have been proposed, but their use in operational forensic science in the United Kingdom is limited due to the perceived complexity with building and populating such models. This paper aims to present an overview of the current method of footwear analysis and interpretation in England and Wales, which can lead to challenges in forensic evaluation, provide examples of where these issues are encountered operationally, and suggest a Bayesian Network model that could address these problems, both in forensic footwear examination and also in broader forensic science practice. A number of models are presented and populated with data from operational databases and with qualitative information provided by an operational forensic footwear expert currently working in the UK. The R v T judgment is used as an example of how such a tool could help to deliver more transparency when communicating forensic evidence and outcomes to the wider Criminal Justice System. A sensitivity analysis of data from different sources is shown, to demonstrate the potential range of values that could be reached for the likelihood ratio and a suggestion of how this could be presented as upper and lower bounds is made.