Les Ainsworth, CRA-Synergy
Most human reliability assessment methods are based upon a task classification scheme which groups together tasks, commonly known as ‘Generic Tasks’, on the basis of some generally functional or psychological characteristics that they share. By various means, quantitative estimates of task reliability are derived for these Generic Tasks and these are then applied as a benchmark that can be applied to other tasks that also share the same basic fundamentals. Clearly, the use of a single probability to indicate the reliability of the disparate range of tasks and situations that can all be classified as a specific Generic Task is a gross simplification, because task performance depends not just upon these features of the task classification, but is also the product of the context in which a task is undertaken. Therefore, in order to more accurately model the reliability that humans can successfully undertake various tasks, the basic Generic Task reliability is adjusted for the impact of various factors such as the adequacy of the human interfaces or the impact of social factors. These factors, are given various names, such as ‘Performance Shaping Factors’ (PSFs) and they are commonly applied as multipliers to adjust a basic Generic Task reliability figure.
Different human reliability methods have adopted different PSFs and even when seemingly similar PSFs are used, their impact upon reliability can differ because they are derived from different data sets. This is not just because PSFs were based upon different selections of studies, but there can also be differences because of the nature of the data sources that were used, which could be based upon laboratory experiments, simulator studies, actuarial data or even expert judgements. Furthermore, it is often difficult to establish where the data came from. Some of the data are also now based upon sources that would now be considered as somewhat dated, so that for instance, data from experimental studies of reading analogue dials are now being used to assess the reliability of using computer interfaces.
In order to assess the effects of using PSFs from different sources, this paper compares some selected PSFs that are ostensibly common to some of the most popular human reliability techniques to identify the differences between them. This should enable weaknesses in the underlying data sources to be identified and rectified and should also permit human reliability assessors to make more informed decisions in their selection of PSFs. It would then be possible to extend this limited set of PSFs and make it available to use with the Generic Task types from any human reliability method.