Abstract
The U.S. Nuclear Regulatory Commission and Sandia National Laboratories sponsored a project to evaluate psychological scaling techniques for use in qenerating estimates of human error probabilities. The project evaluated two techniques: direct numerical estimation and paired comparisons. Expert estimates were found to be consistent across and within judaes. Converaent validity was good, in comparison to estimates in a handbook of human reliability. Predictive validity could not be established because of the lack of actual relative frequencies of error (which will be a difficulty inherent in validation of any procedure used to estimate HEPs). Application of expert estimates in probabilistic risk assessment and in human factors is discussed.
Get full access to this article
View all access options for this article.