To break or not to break? Study on the window-breaking behavior of passengers in subway stations in immersive virtual reality
Can Huang,Jun Zhang,Weisong Liu,Weiguo Song
DOI: https://doi.org/10.1016/j.ssci.2024.106562
IF: 6.392
2024-06-01
Safety Science
Abstract:Passengers trapped in subway carriages could break windows to escape in emergencies, which may be suspected of damaging public properties sometimes. Understanding the decision-making process of passengers and the underlying reasons in such situations is essential for devising specialized management training programs and formulating thorough emergency response strategies. Motivated by this gap, this study investigated how smoke, fire alarm, burnt smell, crowd state, and carriage occupancy influence the decision-making process and response time of passengers with immersive Virtual Reality (VR) experiments. In the experiments, participants needed to decide as in real-life whether to await external intervention or to break the window in an emergency. The proportional hazards duration model, a class of econometric models, was employed to quantify the effects of these factors on the window-breaking behavior and their possible interactions. Combined with questionnaire results, it indicates that smoke, fire alarm, burnt smell, disorderly crowd are significant risk factors, while orderly crowd and occupied carriage are significant protective factors. No apparent linear correlation existed among these risk factors. Nevertheless, analysis of interactive effects demonstrated significant non-linear relationships among certain covariates, which emphasizes the importance of considering interactive effects among factors. It is also observed that participants interpreted psychological cues from occupied carriage differently, potentially influencing their decision-making in emergency situations. Results provide insights for the development of management strategies and the empirical models can serve as a foundation for evacuation models.
engineering, industrial,operations research & management science