Manual vs. automatic control of a supervisory control system

In recent years, a number of studies have examined failure detection performance as a function of the subject's participatory mode (manual vs. automatic mode). The generalizeability of these studies has been limited by both the simplicity of the tasks and the fact that, in general, only failure detection has been examined. The present experiment used a yoked-control experimental design to compare the manual and automatic modes in a full task simulation. The 24 subjects who participated in the study interacted with the system as either a manual controller or as an automatic monitor. Dependent measures were obtained both before and after a system emergency and included total time to complete the task, the number and length of feedback requests, the speed and accuracy of responses to system state questions, and a number of task performance measures. The results indicated superior performance by the manual controllers for a number of the dependent measures. In general, the manual controllers performed the task more efficiently. Although the low fidelity and the simplicity of the task reduced the number of significant variables, the yoked-control experimental design and full-task simulation employed in this study should significantly aid future research.