A logit-based approach for quantifying anchoring effects in evacuation exits choice

Abstract

Understanding how cognitive biases shape evacuees’ decisions is essential for improving building safety. This study investigates the anchoring effect in exit choice under emergency conditions through virtual reality experiments involving 875 participants. Three exit attributes were introduced as prior anchoring information to test their influence on subsequent decisions. In the anchoring information scenario , two exits differed markedly in a attribute where one exit held a relative advantage. This contrastive setup established an initial cognitive anchor for participants before they made choices in task scenarios. A Logit-based discrete choice model was developed to quantify bias magnitude within the random utility framework. Results demonstrate that anchoring information significantly affects exit preferences. The proposed modeling framework advances the quantitative understanding of evacuation behavior and provides actionable insights for enhancing crowd management strategies.