Studying Behavioral Metrics to Design Automatic Intervention Systems in Virtual Environments

Sungjin Hong, Heesook Shin, Youn-Hee Gil, Gun A. Lee

Abstract: In virtual reality training, timely interventions can enhance the effectiveness of education, while poorly timed interventions can reduce training effectiveness. This paper explores methods for determining effective intervention timing in automatic systems within virtual environments. Two intervention types—time-based and action-based—were tested. Results showed significant differences in elapsed time and the number of actions between groups needing intervention and those not requiring it $\text{(p-value < 0.01}$. Further, the average time per action significantly differed between groups preferring action-based versus time-based interventions $\text{(p-value < 0.01}$, suggesting these metrics are key in designing effective automatic intervention systems.