Virtual Reality in Crisis: How Perceived Risk and Warning Accuracy Shape Behavior in Landslide Scenarios

Virtual reality (VR) offers significant potential for disaster preparedness by simulating dangerous scenarios in a controlled environment. However, little is known about how varying environmental elements, such as lighting (day vs. night) and probabilistic warning accuracies, impact behavioral responses in dynamic landslide scenarios. This study assessed participants' responses under four between-subjects VR conditions that varied the accuracy of the warning poles (95% vs. 70% true positives) with day and night environments, all set in rainy weather with a high landslide probability (80%). Behavioral measures, including Euclidean distance around collisions, the number of collisions, and velocity near collisions, were analyzed. Results showed that perceived risk significantly increased behavioral deviations, such as greater Euclidean distances and erratic velocities, regardless of time of day. While higher accuracy poles (95%) significantly reduced collisions and improved stability, lower accuracy poles (70%) led to diminished trust in the system, increasing collision rates and unsafe driving patterns. However, lighting conditions played a less significant role in shaping responses, highlighting the dominant influence of warning system accuracy over visibility. These findings demonstrate VR’s capability to enhance cognitive engagement and training effectiveness for real-world disaster scenarios, emphasizing the need for deploying highly accurate warning systems in landslide-prone regions.