Suchergebnisse

An apparatus and a method for validating estimates of performance time and reliability against empirical measures of human performance time and reliability are described. Measures of performance time were obtained on five tasks and were correlated with estimates of performance times obtained from eight judges in a previous study. Median observed and estimated performance times were highly correlated (r = .98, p<.01). Estimates of maximum performance time corresponded to the 95th to 100th percentiles of the observed distribution of performance time, but estimates of minimum performance time were high and scattered over the lower percentiles. The significant validity coefficient suggests the feasibility of using estimates of performance time, at least for some simple tasks, in system-analytic models when empirical data are lacking and are too expensive to obtain.

This paper assumes increasing use of analytical models in system design. Some characteristics of such models and requirements for human performance data compatible with them are discussed. Methods of obtaining human performance data for use in design models are considered. The use of expert judges to generate performance measures is reviewed. Two new studies are reported in support of the proposition that expert judgments may offer a practical method of obtaining performance measure with potentially wide application in analytical modeling efforts.