Improving trainees’ performances while under stress using real-time feedback

Abstract

Professionals working in different domains often experience stressful conditions evoked by disasters or crisis scenarios. Regardless of these conditions, they have to perform at high standards in order to preserve safety for themselves, avoid any casualties, and to resolve the overall situation. Stress, however, negatively affects cognitive processes and thereby decreases performances. This doctoral thesis aims to improve professionals’ decisions and performances when working in risk- and stressful situations. To reach this goal, a model-based support system was constructed and evaluated. Based on existing models and theories, a model was created to portray the process of performing under stress. The COgnitive Performance and Error (COPE) model explains how work content influences cognitive and affective factors and how, in turn, these factors affect task performances. This model was evaluated during two experiments. First, in a long-term simulated Mars mission, experimental tasks were executed every two weeks. Before, during, and after the tasks, participants subjectively reported their cognitive and affective measures. Results showed that COPE variables differed when work content varied, indicating that work content indeed influenced cognitive and affective variables. Results also showed that task performance could be explained by some cognitive and affective variables. Next, the COPE model was fitted on data collected during a stressful scenario in a high-fidelity Naval simulator. This experiment showed that during virtual training, the COPE-variables were predictors for performances. This resulted in models that could predict performance scores and the number of errors made. The performance predicting models established in the scenario-based Naval training were implemented into a feedback system. This COPE-based FeedBack system (COPE-FB system) provided physiological feedback (heart rate measure), performance prediction feedback, and feedback on the predicted error chances in real-time. When trainees in the high-fidelity Naval simulator received feedback from the COPE-FB system, the amount of errors regarding the planning and speed of task execution decreased. In a laboratory study, participants were confronted with different combinations of the three feedback types while participating in a stressful fire extinguish simulation. Although performances did not improve when the effect of separate feedback types were analysed, all feedback combinations as a whole resulted in an increased performance score. Another result from this study illustrated that participants preferred to receive only the physiological feedback. The studies in this thesis show that the COPE model can be translated into predictive models that use real-time variables to predict performances. Implementing such models into a feedback system resulted in a feedback system that decreased errors in a scenario-based Naval simulator training. In a low-fidelity laboratory study, all feedback combinations in one factor increased overall performance scores. This thesis shows that the COPE-FB system increases parts of trainees’ performances in stressful virtual environments. It also gives some suggestions on how the system can be improved to further increase the trainees’ performances.