Abstract Advanced driver distraction warning (ADDW) systems based on driver monitoring systems (DMSs) must balance safety benefits against driver nuisance. Current distraction guidelines largely do not consider context even though crash risk depends strongly on the available margins in traffic (e.g., time gap). This study investigates whether drivers adapt off-road glance behavior to time-gap–dependent risk during steady-state highway car-following, and whether any such adaptation meaningfully compensates for the increased objective crash risk as time gap gets shorter. Using data from a large realistic (semi-naturalistic) on-road data collection, the eyes-on-road/eyes-off-road behavior was studied in relation to the time gap in car-following scenarios. The glance behavior was extracted from a DMS system, and data about the surrounding traffic were collected by the vehicle’s external sensors. Steady-state car-following segments were extracted and (together with the DMS data) used to estimate off-road glance duration distributions for individual time gaps, including analyses of both overall glance behavior and the longest glance within a segment as an indicator of the drivers’ risk-taking behavior. To quantify safety implications, time-gap–specific glance distributions were combined with a simulation-based counterfactual rear-end crash-risk estimation framework to estimate crash risk. Results indicate only modest behavioral adaptation to increasing crash risk for more safety-critical scenarios: drivers shorten off-road glances slightly as time gaps decrease, with changes most visible in the longest glances, but by far not enough to offset the rapidly increasing crash risk at short gaps. These findings motivate context-aware ADDW strategies in advanced driver assistance systems.
Olleja et al. (Mon,) studied this question.