The power and sensitivity of four core driver workload measures for benchmarking the distraction potential of new driver vehicle interfaces

McDonnell, AS; Imberger, K; Poulter, C; Cooper, JM · 2021 · publications_jsonl

DOI: 10.1016/j.trf.2021.09.019

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates the cognitive distraction caused by voice-based interactive technologies in vehicles, addressing the gap in understanding how auditory/vocal tasks impair driving safety despite allowing drivers to keep their eyes on the road and hands on the wheel. Motivated by the National Highway Traffic Safety Administration’s development of guidelines for in-vehicle interfaces and prior findings that synthetic speech can induce high cognitive workload, the research aims to quantify the mental workload of various voice interactions. Specifically, it examines the sources of workload in speech-based email/texting (comprehension vs. production, natural vs. synthetic voice), the impact of voice commands for vehicle controls, the reliability of menu-based navigation systems, and the cognitive demand of natural language interfaces like Siri. The researchers conducted three controlled experiments involving 45 participants in Experiment 1 (baseline, no driving), with subsequent experiments using a high-fidelity driving simulator and an instrumented vehicle. Participants performed nine nine-minute conditions: a single-task baseline, issuing simple voice car commands, listening to email/text messages read by natural or synthetic voices, listening and composing replies to such messages, interacting with high- and moderate-reliability menu-based navigation systems, and using a hands-free Siri interface for messaging and social media updates. Cognitive workload was assessed using a combination of primary-task driving performance, secondary-task measures (Detection Response Time task), subjective ratings (NASA Task Load Index), and psychophysiological indices (EEG and heart rate). The design allowed for a direct comparison of these voice-based tasks against established benchmarks of cognitive distraction. Results from the baseline assessment demonstrated that not all voice-based interactions carried equal cognitive loads. Detection Response Time increased and sensitivity decreased as task complexity grew, while NASA TLX ratings showed significant increases in mental workload, physical demand, temporal demand, effort, and frustration across more complex conditions. EEG analysis revealed significant effects on P300 amplitude, a marker of cognitive workload, though P300 latency and heart rate did not show significant main effects. The data indicated that activities such as listening to radio or audiobooks were minimally distracting, whereas conversing on a cell phone or using speech-to-text systems for email produced moderate to high levels of cognitive distraction. Notably, the speech-to-text condition with synthetic voice and reply composition generated workload levels significantly higher than traditional voice-based interactions. The study concludes that voice-based interactions, particularly those involving comprehension and production of complex messages or unreliable menu systems, pose significant risks to traffic safety by diverting attention from driving-critical tasks. The findings extend the cognitive distraction rating system, providing a metric that anchors non-distracted driving at the low end and highly demanding tasks at the high end. These results imply that refinements in text-to-speech technology and interface reliability are necessary to reduce cognitive workload, and that current voluntary guidelines may need to account for the substantial mental demands of advanced voice-based interfaces.

Key finding

Task Interaction Time was the most sensitive measure of between-vehicle workload differences, followed by DRT Miss Rate, NASA-TLX, and finally DRT Reaction Time. DRT Miss Rate variance attributable to vehicle ranged from 10% (collapsed) to 20% (navigation), with 80%-power sample size requirements of 10-17 per task. DRT Reaction Time accounted for only ~5% of variance overall (n~89 needed), suggesting cognitive load operated in an on-off fashion at a similar elevated level across vehicles, tasks, and modalities. Correlations between the four measures were weak, indicating each captures a partially unique aspect of workload; Task Interaction Time was largely independent of the other three.

Methodology

on_road

Sample size: N=173 drivers across 40 vehicles and 50 IVIS configurations

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via gog_drive on 2026-06-06.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• workload measurement
• mental demand
• temporal
• visual
• stress driving
• distraction detection algorithms
• time perception

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: behavioral performance data
• Methodological Resource: metric or index, measurement protocol