Exploring occupant exit choices during fire drills and false alarm evacuations in a library

• Published in: Safety science 2025-02, Vol.182, p.106708, Article 106708
• Main Authors: Tong, Yunhe, W. F. Bode, Nikolai, Haghani, Milad, Lovreglio, Ruggiero
• Format: Article
• Language: English
• Subjects: Building; Evacuation; Exit choice; False alarm; Fire drills; Pedestrian dynamics
• ISSN: 0925-7535
• DOI: 10.1016/j.ssci.2024.106708

•497 exit decisions during actual evacuation drills and false alarms in a university library were examined.•Essential factors influencing pedestrian exit choices were analyzed using a multinomial logit model.•Evacuees preferred exits chosen by the majority, nearby, guided by staff, or already familiar.•Crowding and familiarity had less impact on exit choices in drills compared to false alarms. Effective disaster management and public safety rely heavily on understanding human behaviour during evacuations. This study investigates 497 occupant exit choices in real-world evacuation scenarios including two evacuation drills and two false alarm evacuations in a university library building. These authentic settings offer a unique opportunity to examine real-world decision-making processes during evacuations. Employing a multinomial logit model, we quantitatively assess the impact of essential factors on human decision-making. Statistical analysis reveals that participants tend to choose the exits chosen by the majority, closer exits, exits indicated by staff, or exits they are familiar with. We found that participants on the ground floor showed a greater preference for familiar exits compared to those on other floors. Most importantly, we found that in fire drills the effect of crowding and familiarity on exit choices was reduced compared to false alarm evacuations. These findings underscore the critical implications for the conduct of drills and emphasise the importance of studying the contextual dependency of human behaviour during evacuations. Our work also contributes a novel exit choice dataset collected in a naturalistic setting and highlights the importance of the context in influencing pedestrian decision-making during evacuations.