INTENT - A METHOD FOR ESTIMATING HUMAN ERROR PROBABILITIES FOR DECISION-BASED ERRORS

The development of a method, INTENT, for estimating probabilities associated with decisionbased errors is presented. These errors are not ordinarily incorporated into probabilistic risk assessments (PRAs) due to both the difficulty in postulating such errors and to the lack of a method for estimating their probabilities from existing data. By failing to include decisionbased errors in their analyses, most PRA practitioners seriously underestimate the true contribution of human actions to systems failure. This paper attempts to extend the identification of such errors and to quantify them. Two sources, Nuclear Computerized Library for Assessing Reactor Reliability (NUCLARR) and licensee event reports (LERs) were reviewed and two methods, HSYS and SNEAK, were used to identify a generic list of twenty potential errors which may be manifest as erroneous acts. Four categories of influence emerged from the data: consequence, attitudes, response set, and dependency. Corresponding human error probabilities (HEPs) for each error were generated by expert judgment methods. Lower and upper bounds for the HEPs for each error were determined by positing a situation reflecting optimized and degraded performance shaping factors, respectively. To allow analysts the opportunity to refine these extreme HEP values when evaluating a particular scenario of interest, normalization procedures were conducted and generic importance weights were computed for each of 11 performance shaping factors (PSFs) believed to affect the 20 decisionbased errors. It is believed by the authors that PSFs constitute a performance influence which, in some cases, such as in that for training, can serve to either augment or reduce the intellectual resources used by people to successfully accomplish tasks. These derived importance weights are used in conjunction with situation specific PSF ratings to compute a composite PSF score which, in turn, is mapped onto an HEP distribution. Distribution assumptions are presented and a function defining the relationship between composite PSF scores and HEPs is presented for use by the analyst.