Some Prospects for Simulating Human Behavior in High-Rise Building Fires: A Pilot Demonstration

The development of adequate fire safety provisions for buildings is seen to depend upon a valid formulation of a total building fire system, in which human-behavioral and physical phenomena interact. In an effort to comprehend and model such a system, predict human responses in building fires, and assess the usefulness of safety code provisions, a simulation-modeling methodology was evaluated for the case of highrise office buildings. The model investigated, generates human movement decision making behavior under conditions of stress and uncertainty, and is based on the probabilistic Markov process. The internal validity of the present model is examined, by (a) assessing the extent to which variance in the dependent variable (safe egress from the danger zone) is accounted for by predictor variables measured during actual simulation runs, and (b) determining whether the model is capable of distinguishing between diverse spatial designs. Simulation runs were conducted for two different office arrangements, and there were five replications for each arrangement. It was found that (a) Depending on the configuration of predictor variables, up to 88% of the variance in the criterion could be accounted for, and that up to 93.3% of the actual results of simulation runs were predictable by knowledge of these predictors. (b) The present model was incapable of distinguishing between "open-office" versus "compartmentized" designs on the basis of occupants' final egress status, time spent by occupants in threat-invaded zones, or time required by evacuees to reach safety zones. Implications of the findings. and areas for further investigation, are explored.