On Validating Human Performance Simulation Models

Abstract

This paper addresses some of the difficult and elusive problems associated with validating human performance simulation models. Simulation validity can be subclassified into input validity, structure validity and output validity. Of these, output validity is the most objective and also the most important, because it determines whether or not the purpose of the modelling effort can be met. In testing for output validity analysis of variance alone is not sufficient for validating human performance simulation models, as is often taken for granted by many researchers. A more systematic approach is proposed and its implications discussed. The approach is based on considering analysis of variance in terms of the power of the test and a predetermined level of acceptable differences between model and reality.