Citation

Kirwan B. Safety Sci. 1997; 27(1): 25-41.

Copyright

(Copyright © 1997, Elsevier Publishing)

DOI

unavailable

PMID

unavailable

Abstract

Human reliability assessment (HRA) to a significant extent rests on the premise that the various techniques of prediction can reliably estimate human error probabilities. There may be arguments as to whether such an exercise is theoretically valid or even feasible, but many of these arguments are unresolvable at this stage, and since HRA is overall an applied approach, its veracity is best checked empirically. Such investigations of HRA techniques therefore require validations, wherein the outputs of the technique are compared to some standard, to see if the predictions were correct or not. This first paper in a series of two firstly and briefly introduces nine HRA techniques that have been the subject of a number of validations, using various criteria, in the past fifteen years. The paper then derives a common set of criteria with which to validate techniques. Such criteria can then be used uniformly across validations to decide if a technique demonstrated empirical validity or not. The second objective of the paper is to classify validation exercises themselves, in terms of the quality of data and subjects utilised in the validation (related to external validity), and the degree of experimental rigour/control adopted in the validation (related to internal validity). Such classifications allow the evidence from individual validation studies to be informally 'weighted', so that validation evidence from different 'levels' of validation can be aggregated. This allows conclusions to be drawn on the validity of these HRA techniques, even where there are apparently conflicting validation results. The criteria and classifications developed in this paper are then utilised in the second paper, which evaluates twenty-two validation and verification studies, and summarises the evidence on the validity of the nine techniques, the need for further validation of HRA techniques, and the need for more holistic validation of HRA.