Predictors of evacuation behavior: dataset on respondents' route choice and web interaction.

Empirical data on human evacuation behavior are invaluable for adjusting and training computational algorithms that simulate evacuation processes, including agent-based modeling. We provide a dataset on human decision-making during evacuations from virtual buildings, captured using experimental methods that controlled specific building layout parameters. An online experiment assigned participants a random subset of tasks featuring T-intersections. Data from 208 respondents, aged 17 to 71, were analyzed, considering education levels and excluding those with significant technical issues. Quantitative data on user interaction and evacuation route choices included decision time, mouse rotation, and the selected corridor, recorded through mouse clicks on invisible areas of interest. Respondents also self-reported their choice confidence on a Likert scale. Additionally, responses to final retrospective evaluation questionnaires were recorded. This dataset offers diverse research opportunities, particularly in emergency evacuation planning, where understanding evacuation choices in simulations can inform real-world strategies. It supports the development of models to predict human behavior in emergencies using machine learning and predictive modeling and is accessible for both academic and commercial use.