Developing a Methodology for Eliciting Subjective Probability Estimated during Expert Evaluations of Safety Interventions: Application for Bayesian Belief Networks Final technical rept

摘要

The NASA Aviation Safety Program (AvSP) has defined several products that will potentially modify airline and/or ATC operations, enhance aircraft systems, and improve the identification of potential hazardous situations within the National Airspace System (NAS). Consequently, there is a need to develop methods for evaluating the potential safety benefit of each of these intervention products so that resources can be effectively invested to produce the biggest benefit to flight safety. Of interest to the present project is the process of using expert judgments to develop Bayesian Belief Networks (BBNs) that model the potential impact that specific interventions may have. Specifically, the present report summarizes methodologies for improving the elicitation of probability estimates during expert evaluations of AvSP products for use in BBNs. The work involved joint efforts between Professor James Luxhoj from Rutgers University and researchers at the University of Illinois. The Rutgers project to develop BBNs received funding by NASA under contract NAS1-03057 entitled 'Probabilistic Decision Support for Evaluating Technology Insertion and Assessing Aviation Safety System Risk.' The proposed project was funded separately but supported the existing Rutgers program.