19 - Modeling and Predicting Human Error

Abstract

This chapter discusses the modeling and prediction of human error. Error models cannot be surveyed and rated for inclusion in a computer-aided engineering and design framework because a model of error is also a model of processing mechanisms and, in part, because there are few models available that address the way in which errors occur at the scale of behavior relevant to pilot performance. The models of processing mechanisms are not necessarily models of the way in which processing can break down or lead to erroneous performance. The error models available are either descriptive taxonomies or cognitive simulations that assist an analyst in discovering error-prone points in a person-machine system. There have been long and unresolved debates among researchers on human performance as to what human error is. The concept is to separate performance failures from information-processing deficiencies. Performance failures are defined in terms of a categorical shift in consequences on some dimension related to performance in a particular domain. A bridge is needed between the manner in which processing may unfold and the domain consequences of that processing.