An Evidence-based Cognitive Model of Uncertainty during Indoor Multi-level Human Wayfinding

Existing computational models lack adequate representation of the uncertainty experienced in human wayfinding tasks. They overly rely on optimized pathing algorithms, which reduces realism and limits insights on human responses to architectural designs. To address this, we developed an empirically grounded model that predicts human wayfinding uncertainty experience. Using data from 28 participants navigating an educational building with varying signage, we constructed the model (Study 1), and validated it with data from 11 other participants (Study 2). We found the wayfinding uncertainty correlated with the time elapsed since seeing the last helpful sign. The cognitive agent based on this model closely replicated human-reported uncertainty levels during wayfinding tasks under different signage conditions. Although the model more closely resembled human behavior compared to a shortest-route algorithm, additional environmental variables and heuristics are needed for better human outcome alignment. Our study showcases that evidence-based cognitive agent modeling can provide nuanced, human-like wayfinding behavior, enhancing the potential for effective computational design evaluation.