Modeling Microstructure of Drivers' Task Switching Behavior and Estimating Crash Ris

Driver distraction has been a longstanding cause of vehicle crashes. To understand and predict distracted drivers' task switching behavior between driving and secondary task and to mitigate negative consequences of distraction associated with inappropriate task switching, researchers have analyzed glance behaviors and developed models of driver behavior. In most of the studies, however, the microstructure of glance patterns that shows the process of drivers' engagement and disengagement with secondary task has been neglected.;This dissertation presents a computational model of driver distraction that accounts for the joint and dynamic influence of external factors (influence of uncertainty and task structure) and internal factors (individual differences), which has been not well demonstrated in previous driver models. The model can predict drivers' glance patterns associated with a secondary task, making it possible to quickly evaluate potential danger of many candidate task designs and to better understand cognitive mechanisms underlying distraction. The model specifically focuses on the effect of task structures (i.e., subtask boundary designs) on dynamic task switching behavior. The model shows how subtask boundaries influence how frequently and how long drivers---each with different characteristics---interact with secondary tasks.;Another gap of the current literature is that the actual crash risk that different task structures contribute to driver distraction has not been extensively studied. This dissertation also estimates crash and injury risk associated with different subtask boundary designs, using a counterfactual simulation, which explores how alternative glance patterns might alter the observed consequence of an event.;Overall, the dissertation spans an empirical study (Chapter 3), computational modeling of task switching behavior (Chapter 4), and simulation of crash risk (Chapter 5). This set of studies explores the effect of task structures on glance patterns and quantifies the risk of driver distraction caused by different task structures. The model's approach of inferring cognitive mechanisms (micro-level) underlying the overall system outcomes (macro-level) is rigorous and useful in understanding the overall cost or benefit of task switching, not only in driving situations but also in other multitasking situations with time-critical tasks. The computational model and simulation may also be used to assess how a task affects multitasking performance in the early stage of task design.