Simulator Fidelity and Validity in a Transfer-of-Training Context

Abstract

The question of simulation fidelity and validity is of concern in many driving applications. This paper reviews the general issues associated with fidelity and validity. For training applications, fidelity and validity relate to training effectiveness. Training effectiveness is typically measured according to real-world driver behavior and performance. This paper describes a novice driver training experiment that used simulators with three levels of fidelity. Driver simulator performance was measured for complex driving scenarios containing repeated exposures to hazardous situations. The driving records of the participants were also analyzed for a period of 40 months after they were trained. Inferences about simulator fidelity and validity are drawn on the basis of the participants' simulator performance and real-world driving records. Confounding effects in the database due to several uncontrolled variables are also discussed. These confounding effects are relevant to the design of future transfer-of-training studies.