A Secondary Assessment of the Impact of Voice Interface Turn Delays on Driver Attention and Arousal in Field Conditions

Abstract

Summary:Voice interface use has become increasingly popular in vehicles. It is important that these systems divert drivers' attention from the primary driving task as little as possible, and numerous efforts have been devoted to categorizing demands associated with these systems. Nonetheless, there is still much to be learned about how various implementation characteristics impact attention. This study presents a secondary analysis of the delay time between when users finish giving commands and when the system responds. It considers data collected on 4 different production vehicle voice interfaces and a mounted smartphone in field driving. Collapsing across systems, drivers showed an initial increase in heart rate, skin conductance level, and off-road glance time while waiting for a system to respond; a gradual decrease followed as delays continued. The observed attentional and arousal changes are likely due to an increase in anticipation following a speech command, followed by a general disengagement from the interface as delay times increase. Safety concerns associated with extended delay times and suggestion of an optimal range for system response times are highlighted.

Multi-modal demands of a smartphone used to place calls and enter addresses during highway driving relative to two embedded systems

“…While the smartphone largely outperformed the Volvo Sensus voice system in task time and total off-road visual engagement for the destination address entry task, it showed more modest improvements in comparison to the Chevrolet MyLink system where predominant difference related to an improvement in task time. The degree to which these differences relate to differences in task structure, display characteristics or other variables such as system response time (McWilliams et al 2015), voice recognition system performance, ambient noise, etc. are all potential contributors that need to be actively optimised at the system design level.…”

Additional Findings on the Multi-Modal Demands of “Voice-Command” Interfaces

“…The first study in the MKS [1] also considered a voice-based song selection task. The primary comparison in the present report is to the second study in the MKS [4] since the NHTSA age distribution was used in both, whereas the first MKS study considered a younger (20)(21)(22)(23)(24)(25)(26)(27)(28)(29) and relatively older (60-69) sample. The present study in the Impala again considered voice-based full address entry into the navigation system and the nback auditoryvocal reference task.…”

“…The time delay was significantly longer in the Impala than what was experienced in the Equinox, XC60, and MKS. An assessment was carried out looking at delay times across the four vehicles and a smartphone voice interface; this assessment has been developed as a separate paper [20]. Mean delay times assessed in the Impala averaged around 8 seconds and, in some instances, exceeded 15 seconds.…”

“…The demands on attention associated with using production voicecommand based interfaces under actual driving conditions for destination address entry and POI selection can be quite substantive in a number of respects (self-reported workload, TEORT) yet fall well within what might be considered reasonable limits for others. As noted, the impact of factors such as turn delay [20] on an otherwise well designed DVI should be taken into account, as well as the impact of frequency of errors and provisions for straightforward error recovery. Present implementations of these DVIs are clearly worthy of continued analysis to further identify design features and principles that best support functional attention to the primary driving task.…”