Estimating parameter uncertainty bounds of human error probability using Monte Carlo simulation

Abstract

Characterizing the uncertainty of reliability in risk assessment has been recognized as a critical element in effective decision-making. Although many kinds of risk assessment guidelines have introduced various uncertainty analysis techniques, estimates of the uncertainty bounds of human reliability have not been investigated on an empirical basis. In this study, we predicted how parameter uncertainties in human error probabilities can be formed by employing uncertainty intervals in the failure probabilities of primitive tasks, the multipliers of performance-shaping factors, and recovery failure probabilities. The above component values were incorporated through Monte Carlo simulation and the EMBRACE (Empirical Data-Based Crew Reliability Assessment and Cognitive Error Analysis) method. As a result, distributions of human error probabilities in various contextual situations were derived. The statistical model most suitable for the distributions was then selected, and the bounds of 90 % uncertainty intervals were estimated according to the selected statistical model. Implications and limitations of this study are also discussed.