It's hard to resist the temptation to do two things at once, even when one of those things is driving. To make multitasking easier and safer, automakers and technology companies have provided drivers with the ability to use voice commands to operate smartphones and infotainment systems.
Voice systems do help drivers keep their eyes on the road when compared with manual interfaces, but they don't eliminate visual distraction altogether, a new study by researchers from IIHS and the Massachusetts Institute of Technology's AgeLab shows. And when researchers compared embedded voice systems from two vehicle makes and the voice interface of a smartphone, they found considerable differences among the three in terms of the visual demand, time involved and accuracy.
"In an ideal world, drivers wouldn't do anything but drive while the vehicle is moving. But people are increasingly plugged in at all times, and automakers have responded by installing systems to make it easier to use technology on the go," says Ian Reagan, an IIHS senior research scientist and a co-author of a pair of papers based on the experimental study. "While you can't completely eliminate the distracting nature of these types of tasks, this study shows it's possible to reduce some types of distraction through system design."
The study was conducted on interstates in the Boston area with 80 participants ages 20-66. Half of the group drove a 2013 Chevrolet Equinox equipped with the Chevrolet MyLink system, and half drove a 2013 Volvo XC60 with the Volvo Sensus system. All participants used a Samsung Galaxy S4 smartphone programmed with the same list of more than 100 contacts and mounted in the center console area.
MyLink and Sensus were chosen for the experiment after an earlier IIHS study revealed big differences between them. In that study, researchers counted the steps involved in calling a contact via four vehicle infotainment systems. Calling a contact using voice commands with Sensus required multiple statements to navigate through different system menus, while the same task required a single detailed voice command with MyLink. The systems' visual-manual interfaces also differed. To call a contact manually, Sensus required the driver to scroll through the contact list using a rotary knob, while, with MyLink, the driver used a rotary knob and push-button to access the alphabetical range containing the desired contact and then scroll through that more limited list.
In the Boston experiment, each driver was trained in the use of the vehicle system while parked and then had to use the system to complete a series of tasks while driving: calling a contact manually, calling a contact using voice commands and entering an address for navigation using voice commands. The drivers also were trained in use of the Galaxy S4 and completed the same tasks using the smartphone directly.
Researchers later used video footage and vehicle performance data to analyze off-road glances, task completion time, errors, vehicle speed, speed variability and steering wheel reversal rates. EKG and skin conductance sensors were used to measure the drivers' physiological responses, and drivers also rated the workload involved in each task. Participants were told "workload" could include both mental and physical effort, as well as things like distraction and frustration.
The participants were instructed to give priority to safe driving. Before the driving began, eight people were excluded from the study: two because one of the systems couldn't interpret their speech, two who said they were uncomfortable performing the tasks while driving and four who had difficulty learning how to use the systems. Three participants were withdrawn because of safety concerns about their driving. None of these are included in the final count of 80.
None of the voice interfaces eliminated all glances away from the road when calling a contact, but all of them reduced total eyes-off-the-road time to some extent compared with manual calling. When using the smartphone with voice commands, drivers looked away for a total of 13 seconds on average, compared with 15 seconds when dialing a contact manually. The reduction was more substantial for both of the embedded systems, particularly MyLink. The Chevrolet system required an average of 14 seconds of off-road glances for the manual interface and 3 seconds for the voice interface.
The picture was more complicated when it came to entering an address. Drivers looked away from the roadway for much more time with the Sensus system than with either MyLink or the smartphone. That's because Sensus' menu-based design required each element of the address to be entered separately and allowed the driver to look at a center-mounted display to verify that the previous component was interpreted correctly and fix it if necessary. While the design had higher visual demand, it resulted in relatively few errors.
In contrast, drivers using the MyLink system had to give the complete address all at once. This "one-shot" system was much quicker than the menu-based approach, but many drivers in the study had trouble getting the system to understand the address correctly or made mistakes as they recited the address in a single string. Researchers in the car noticed that some participants appeared to try to help the system by speaking slowly, but this only seemed to confuse it more. Out of 120 attempts to enter an address via MyLink, 38 had system errors, and 23 had user errors. In contrast, there were only five system errors and eight user errors with Sensus.
Voice interface design:
Two approaches to entering an address
Volvo's voice system required more steps to enter an address than Chevrolet's but gave drivers the chance to correct errors along the way.
"The one-shot approach of MyLink's voice interface seemed to work well for contact calling, but a full address may be too complicated for that method," says David Kidd, an IIHS senior research scientist and study co-author. "A high error rate could negate some of the benefits of fewer off-road glances. When drivers become frustrated with technology, that itself can be distracting. Or they might give up on the system and resort to another, potentially more distracting navigation method."
In addition to the one-shot approach, another possible reason for the high error rate could be the fact that the Equinox had more ambient noise than the XC60. That could help explain why the smartphone, which was used identically in both vehicles, had a higher rate of system errors when used for voice-based address entry in the Chevrolet than in the Volvo.
"Would MyLink have performed better in a quieter vehicle?" Kidd asks. "That's an open question."
The wide variety of both smartphones and vehicle systems makes it difficult to generalize from the study, and the outcome might have been different if participants had been familiar with both the phone and the vehicle. What's clear, however, is that voice input has some benefits compared with manual input, and there are pluses and minuses in different designs.
When it comes to cellphones and infotainment systems, many safety advocates are concerned that hands-free and voice-activated systems don't eliminate cognitive distraction. In the current study, cognitive distraction wasn't specifically measured. The researchers used participants' self-reports about workload, as well as the physiological indicators and driving performance measures, to try to gauge how demanding the tasks were. The drivers rated the voice interfaces as less demanding than manual ones, but the voice interfaces still led to modest changes in driving performance and increases in stress levels relative to periods when drivers were just driving.
"Cognitive distraction is a real concern but a difficult one to study," Reagan says. "However, regarding visual distraction, there's no question that if you're at least looking at the road ahead, you have a better chance of not crashing into something in front of you than if you're looking at a vehicle infotainment display or at your cellphone."
Cognitive distraction can result from all kinds of behaviors, many of which would be impossible to eliminate, such as talking to passengers or daydreaming. Crash avoidance technologies that can prevent or mitigate impending crashes, no matter the cause, represent a promising solution to this problem, and more and more vehicles are being equipped with such systems (see Status Report special issue: distracted driving, Oct. 24, 2014).