Developing a Decision Analytic Framework Based on Influence Diagrams in Relation to Mass Evacuations

In this paper, we examine the role which decision analysis can play in a situation requiring a mass evacuation. In particular, we focus on the influence diagram as a tool for reasoning and supporting decision-makers under conditions of risk and uncertainty. This powerful modelling tool can help to bridge multiple specialist domains and provide a common framework for supporting decision-makers in different agencies. An influence diagram is also referred to as a decision network and can be considered as an extension of a Bayesian network. Like a Bayesian network, it contains chance nodes which represent random variables and deterministic nodes which represent deterministic functions of input variables. However, in addition, an influence diagram contains decision nodes which represent decisions under local control and utility nodes which can represent a variety of costs and benefits. These might be measured in several dimensions including casualties and monetary units. Advantages of Bayesian networks and influence diagrams over more traditional risk and safety modelling approaches such as event trees and fault trees are discussed - in particular, the ease with which they represent dependencies between many factors and the different types of reasoning supported at the same time, e.g. predictive reasoning and diagnostic reasoning. An illustrative, generic influence diagram is presented of a situation corresponding to a CBRNE attack. We then consider how this generic model can be applied to a more specific scenario such as an attack at a sporting event. A variety of potential uses of the model are identified and discussed, along with problems which are likely to be encountered in model development. We argue that this modelling approach provides a useful framework to support cost-effectiveness studies and high-level trade-offs between alternative possible security measures and other resources impacting on response and recovery operations.