Aviation causal model using Bayesian Belief Nets to quantify management influence

The authors have recently completed a causal model for aviation safety under contract with CAA the Netherlands and the US Federal Aviation Administration. The goal was to develop a comprehensive model of aviation safety and to judge the potential impact of management decisions. The overall modeling tool chosen for this task was Bayesian Belief Nets (BBNs). The reason for choosing BBNs above more conventional tools (fault and event trees) were: 1) BBNs better enable the problem owner to recognize his problem; 2) Software support is such that the top level representation of the problem is at once the user interface for performing cal- culations; 3) The influence of management decisions is readily factored in using structured expert judgement. Perhaps the most satisfying aspect of the BBN approach is that it provides a comprehensive model which is maximally data-driven, yet which includes assessments of potential impacts of contemplated decisions.