Learning to predict human error: issues of acceptability, reliability and validity

Abstract

Human Error Identification (HEI) techniques have been used to predict human error in high risk environments for the past two decades. Despite the lack of supportive evidence for their efficacy, their popularity remains unabated. The application of these approaches is ever-increasing, to include product assessment. The authors feel that it is necessary to prove that the predictions are both reliable and valid before the approaches can be recommended with any confidence. This paper provides evidence to suggest that human error identification techniques in general, and SHERPA in particular, may be acquired with relative ease and can provide reasonable error predictions.