In-Vehicle Voice Control Interface Evaluation: Preliminary Driver Workload and Risk Analysis

Jenness, James; Boyle, Linda Ng; Lee, John D.; Miller, Erika; Yahoodik, S.; Huey, Richard; Lee, Ja Young; Benedick, A.; Petraglia, Elizabeth · 2020 · ROSA P / United States. Department of Transportation. National Highway Traffic Safety Administration

archive: archived pipeline: cataloged verified

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

Summary

This report evaluates the safety risks and driver workload associated with in-vehicle voice control systems (VCS). Sponsored by the National Highway Traffic Safety Administration, the research aims to establish empirical measures for cognitive and visual distraction and link these metrics to crash risk. The study addresses the need for standardized evaluation protocols for auditory-vocal interfaces, which are increasingly common in vehicles but lack comprehensive safety guidelines compared to visual-manual tasks. The project comprised three interconnected studies. Study 1 utilized a driving simulator with nine participants to assess a dual detection response task (DRT) protocol. Researchers compared the standardized Tactile DRT (TDRT), which measures cognitive load, with a modified Remote Visual DRT (RDRT) designed to be sensitive to visual attention away from the roadway. Participants performed various VCS tasks, including audio-only and hybrid (audio-visual) navigation and radio tuning, while performing these DRTs. Study 2 compared the robustness of evaluation measures across contexts by having the same participants perform tasks in both a driving simulator and on-road conditions. Metrics included TDRT response times, speed matching to a lead vehicle, brake light response times, and eye glance behavior. Study 3 employed counterfactual simulation to estimate relative crash risk. Researchers substituted the eye glance data collected in Studies 1 and 2 into a dataset of 34 crashes and 190 near-crashes from the SHRP 2 naturalistic driving study to simulate "what-if" scenarios regarding crash outcomes. Key findings indicate that response time was sensitive to differences between VCS tasks, with a significant interaction between DRT type and task complexity. Hybrid VCS tasks, which included visual displays, were completed more quickly and accurately than audio-only tasks, as visual displays allowed drivers to control information pacing. Consequently, audio-only tasks were perceived as more demanding. Eye glance data revealed that total eyes-off-road time was generally higher for TDRT protocols than for the modified RDRT, suggesting the modified RDRT encourages forward visual attention. In Study 2, TDRT and eye glance measures proved robust across simulator and on-road contexts, whereas speed correlation and brake light response times were less reliable or failed to distinguish tasks from baseline driving. Study 3 demonstrated that counterfactual simulation could differentiate relative crash risk by task and driver, though absolute risk estimates depended heavily on the severity of the input crash scenarios. The significance of this work lies in its development of a framework for linking VCS evaluation metrics to safety outcomes. The study validates the use of TDRT and eye glance measures as robust indicators of distraction across different testing environments. Furthermore, it highlights that hybrid VCS interfaces may impose less cognitive burden than audio-only systems. The counterfactual simulation method offers a promising tool for estimating relative risk, although the authors caution that results should be interpreted as relative rather than absolute indicators due to dependencies on the specific crash event samples used.

Key finding

Detection response times and eye glance measures effectively distinguished between voice control task demands and remained consistent across simulator and on-road contexts, while counterfactual simulations demonstrated that relative crash risk estimates varied significantly by task and driver.

Methodology

mixed_methods

Sample size: 9

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 bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).

StageOutcomeToolModelPromptAttemptsCompleted
discover success rosap 2 2026-05-23
archive success 1 2026-05-23
extract success cached 2 2026-06-10
clean success 1 2026-06-01
chunk success 1 2026-06-01
embed success 1 2026-06-02
enrich success 1 2026-05-23
promote success 1 2026-05-23
summarize success llm qwen3.6-27b-prismaquant summ-v5 3 2026-06-10
tag success vector_similarity 19 2026-06-11
verify success 2 2026-06-10

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.

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).