A Causal Model of Intersection-Related Collisions for Drivers With and Without Visual Field Loss

Causal models allow us to reconstruct traffic accidents, predict the likelihood of future accidents and implement counter measures to prevent them. For drivers with impairments like visual field loss, naturalistic data on crash causes is however scarce due to their current prohibition to drive. This paper presents an approach to deriving a causal model for the prediction of crash risks for current non-drivers. The applied use case focuses on a collision with an overlooked crossing vehicle in an intersection. Based on the combination of crash analyses for normal sighted drivers and models of information processing and human errors, a general structural causal model for crash risks in this use case was developed. The application of this model to drivers with visual field loss on the side of the approaching vehicle revealed four causal factors with an increased risk of occurring: faulty anticipation of location and timing of hazards; inadequate guidance of gaze movements; adverse scanning patterns; and cognitive overload. These elevated crash risks can guide the development of assistive technologies for drivers with visual impairments in the future.