Inferring Hidden Attentional States in Driving: A Bayesian Approach to Modeling Distraction and Secondary Task Engagement

Objective. To develop and validate a computational framework that infersindividualized attention strategies and latent distraction states to support personalizedmodeling of multitasking behavior and intervention.Background. Driver distraction from in-vehicle systems is a growing safetyconcern. However, the level of distraction is often latent and varies significantly acrossindividuals. Existing models typically overlook these differences, limiting their effective usefor personalized interventions.Method. We introduce a Partially Observable Semi-Markov Decision Process(POSMDP) to model hidden attentional dynamics and attention allocation decisions. Usingbehavioral data, including glance behavior, velocity, and pupillometry, from a high-fidelitydriving simulator with 18 participants, we estimate personalized reward functions thatreflect each driver’s subjective valuation of secondary task utility versus safety cost.Results. The method accurately infers distraction states and recoversparticipant-specific utility weights governing the trade-off between secondary task benefitand driving cost. Compared to a well-established 2-second glance rule, it improvesdetection of distraction events and reveals individual variability in attention strategies.Some drivers exhibit highly conservative profiles, while others assign greater value tosecondary tasks, even under high distraction. Counterfactual simulations show howperceived task importance could modulate visual attention behavior across individuals.Conclusion. Our POSMDP-based framework provides an interpretable andindividualized account of driver attention allocation, capturing both latent states andbehavioral variability.Application. This model enables the development of personalized, risk-sensitivedriver assistance systems that adapt to individual attention strategies, enhancing roadsafety through context-aware, graded interventions.