Methods for including human variability in system performance models

Abstract

To understand system performance, it is rational to consider all system components, including the humans involved in the control or maintenance of the system. Previous research has included human performance by modeling human tasks as events within Discrete Event Simulation (DES) models. These models typically represent the variability of task performance times and error rates by calculating the mean and variance across multiple individuals. Such approaches assume independence of task performance measures between individuals, but evidence exists which indicates that task performance measures are correlated between individuals. The current research seeks to understand methods to account for performance variability within DES models. A taxonomy of potential methods to address variability in DES models is developed and discussed. Among the findings derived through development of this taxonomy is the need to differentiate models of performance envelopes from models of average system performance and alternatives for modeling the human when predicting each class of performance.