In-Vehicle Interface Adaptation to Environment-Induced Cognitive Workload
URL: http://arxiv.org/abs/2210.11271
archive: archived pipeline: in_review verified
Abstract
Many car accidents are caused by human distractions, including cognitive distractions. In-vehicle human-machine interfaces (HMIs) have evolved throughout the years, providing more and more functions. Increased HMI functions lead to increased cognitive workload, which can be hazardous when combined with environmental conditions. We propose an HMI adaptation framework that decreases the user's cognitive load by adjusting interface components based on the environment-induced cognitive workload.
Summary
Driving-simulator user study (OpenDS) testing whether an adaptive in-vehicle HMI that simplifies its display in visually complex driving environments reduces driver mental workload (MWL) and improves user experience compared to a static HMI. 35 drivers (16 adaptive, 19 static, between-subjects) drove a ~30-min route from a low-complexity countryside to a high-complexity city environment while performing easy / medium / hard / no-task secondary menu interactions on a phone/navigation/media interface. MWL indexed by heart rate (HR) and heart-rate variability (HRV) via Polar H10 chest sensor, plus driving performance, secondary-task latency/success, and UEQ+ user-experience ratings. Four hypotheses tested adaptive vs static differences and task-difficulty interactions.
Key finding
Contrary to predictions, the adaptive HMI did not reduce MWL relative to the static HMI: the static group actually showed a smaller increase in clicks and a larger learning effect from countryside to city, HR decreased (rather than increased) in the city condition suggesting an ineffective workload manipulation or compensatory speed reduction, and UEQ+ ratings did not differ between groups. The authors conclude that interface adaptation can confuse drivers and that more careful manipulation of environmental difficulty plus driving-experience controls are needed.
Methodology
Between-subjects user study in a medium-fidelity driving simulator (OpenDS). 35 participants assigned to adaptive (n=16) or static (n=19) condition. Two driving environments (low-complexity countryside vs high-complexity city) crossed with four secondary-task segments (no-task, easy, medium, hard) of phone/navigation/media menu interactions; static condition kept the complex interface in both environments while adaptive switched to a simplified interface in the city. Measurements: HR and HRV from a Polar H10 ECG chest sensor; longitudinal and lateral driving-performance metrics; secondary-task latency, success count, and relative success; UEQ+ post-study questionnaire. Statistical analyses: one-sample two-sided t-tests on city-minus-countryside differences for HR/HRV, ANOVA / repeated-measures with task-difficulty as within factor.
Sample size: 35 participants; adaptive group n=16 vs static group n=19 (between-subjects). Static group reported significantly higher annual driving km (M=10639, SD=4259) than adaptive group (M=2733, SD=12835), t(33)=2.35, p=.025.