How biases in urban mental representation affect human navigation

Pedestrian navigation emerges from the complex interplay of topological and perceptional criteria. Whilst topological criteria are objective - perceptional criteria are individual and closely related to how people perceive the urban environment. To isolate both effects, we introduce a simple mobility model that try to reproduce aggregated properties observed in over 200,000 human paths collected through a smartphone app. The model considers the effects of biased perception of pedestrians on streets and the herding effects. To simulate the biased perception in human navigation, we hypothesize that humans aim at minimizing walking distance but, due to bias of the urban environment, they have a limited ability to precisely estimate distances. We model this as random perturbations to the perceived length of street segments while searching for the shortest distance path, controlled by what we call a polarization parameter. After observing that the assumption of imperfect perception of distance is not sufficient to reproduce all the observed aggregate properties of human paths, we make the additional assumption that such imperfect perceptions might be correlated across individuals. I.e., in accordance with existing literature we assume that different individuals might have similar distortions of distance caused by the urban environment, and introduce this into our model through a herding parameter. After the calibration with a set of 223,928 actual human paths in the city of Boston, the proposed model successfully reproduced many of the aggregate statistics observed in human paths, such as length and exploration area. Calibrations of the herding parameter further enabled the model to replicate the predictability of the flows and popularity of road intersections. Our results confirm that the biased perception of pedestrians on the streets could be a major factor that *differentiates human and shortest paths in street networks. The calibrated herding parameter indicates that pedestrians share about 50% of the biased perceptions on street segments, providing evidence to the assumed existence of the herding effect in human perception that could explain the heavy-tailed distribution observed in a wide range of human dynamics variables. To conclude, we demonstrate with a case study identifying human-preferred navigation points in Boston that the proposed model could further benefit future studies on human spatial navigation as a fixed-length null mobility model to help isolate human preferences from topological effects.