VIDEO: MIT and IIHS Suggest Voice Commands Can Reduce Distraction, But Don’t Eliminate It

Posted by

From the very first moment that car phones were bolted to the floor in automobiles, it was obvious that the distraction of dialing the phone was detrimental to driving safety. Now that mobile phones have eclipsed the functions of a normal web browser, it’s become even more important to reduce the overall demands on drivers. The simple answer was thought to be voice commands, but in a joint study from MIT AgeLab and the Insurance Institute for Highway Safety found that the picture may be more complex. While voice commands can reduce some visual demand when entering stored phone numbers or addresses into navigation, but do not eliminate it, while creating the potential for a lot of errors in the process.

The new study authored by Bruce Mehler, Bryan Reimer and Jonathan Dobres from MIT AgeLab and the New England University Transportation Center, and David Kidd, Ian Reagan and Anne McCartt from the IIHS notes “A perceived advantage of voice inputs compared with manual inputs is that they eliminate or reduce the competition for visual and manual resources between a secondary activity and the primary task of driving. Therefore, voice interfaces have been widely considered as an appealing approach for giving drivers access to a range of entertainment and connectivity options while minimizing the potential impact on driving performance and safety.”

The new study assesses driving performance, visual engagement and “indices of workload” such as heart rate, skin conductance and other subjective ratings in 80 drivers.


The drivers were randomly assigned to drive either a 2013 Chevrolet Equinox or a Volvo XC60. The Equinox was equipped with Chevrolet’s MyLink system, while the Volvo featured the Sensus system.


MIT AgeLab’s Bryan Reimer and Bruce Mehler conducted a series of related studies in a 2013 Lincoln MKS using Lincoln SYNC. The study suggested that that compared with manual interaction, voice interaction with embedded and portable systems can reduce visual demand as intended, but do not necessarily eliminate it. In fact, some voice interactions result in “moderate to large visual engagement” when measured in “glance time to device or total eyes off the road time.”

Participants in the latest MIT AgeLab and IIHS study were all from the Boston area, located via newspaper and online advertisements. The participants were equally distributed across several age groups (18-24, 25-39, 40-54, 55 and older).

Participants drove one of two vehicles, either a 2013 Chevrolet Equinox equipped with the MyLink system or a 2013 Volvo XC60 equipped with the Sensus system, each of which paired with a Samsung Galaxy S4 via Bluetooth.


Both vehicles were equipped with data acquisition tools that integrated with the car’s CAN bus, a Garmin 18X Global Positioning System (GPS) unit, a MEDAC System/3TM physiological monitoring unit to provide EKG and skin conductance level (SCL) signals, video cameras, and a wide-area microphone to capture driver speech and audio from the vehicle’s speech system.

The five video cameras captured the driver’s face for primary glance behavior analysis, the driver’s interactions with the vehicle’s steering wheel and center console, the forward roadway (both narrow and wide-angle images), and a rear roadway view.

Selecting two different cars with two different interfaces exemplifies the challenges that car shoppers are faced with when buying a new car. For the most part, three point seat belts and many other safety items are fairly standardized, but voice activation and smartphone integration vary wildly between manufacturers.

Chevrolet MyLink and Volvo Sensus appear to represent the range of voice control options.

Phone Contact Entry – MyLink vs. Sensus

When calling a stored contact with Chevrolet MyLink, drivers use just a single voice command. The driver says “Call Home,” and the system responds “calling Home on cell” as the call is initiated.

The same exact task using Volvo Sensus requires four separate voice commands. To call “Home” from stored contacts, the driver says “Phone, call contact,” then waits for the system prompt “name please.” The driver then responds “Home” and waits for the system prompt, “Please say a line number.” The driver responds with the appropriate line (typically “one,” but it can be any number of lines depending on how many are stored for that contact). The system then responds “Dial Home mobile – confirm: yes or no.” After an affirmative response, Sensus then makes the call.

According to the study, calling a contact with the Volvo Sensus took more steps when using voice inputs than when using manual inputs. “Furthermore, many of the system prompts asked the driver to look at the center stack display to choose among options for the contact (e.g., home, cell, work) or to confirm input, and the prompt asking for a line number always occurred whether the contact (“home’ in the example here) had one or multiple line numbers.”

MyLink appears to be easier to operate, but according to the study, there’s still demand that requires attention. “While calling a contact in the phone book with Chevrolet MyLink required fewer voice commands than Volvo Sensus, MyLink required a deeper understanding of system operation in that it did not provide as much prompting, visual support, or confirmation, which could potentially result in more calling errors.”

Navigation Address Entry – MyLink vs. Sensus

The study also looked at how voice activation influenced demand when entering an address in to the navigation system. Similarly, navigation entry varied significantly between the two systems. Like phone contact entry, MyLink requires fewer steps to enter an address. The driver presses the  “push-to-talk” button and says “Navigation.” MyLink is flexible enough to accept various commands including “destination address,” “enter address,” and simply “address.” At that point, the driver speaks the full address in a single verbal string, in this case: “177 Massachusetts Avenue, Cambridge, Massachusetts.” The system then has to parse that string of addresses into its component parts.

Similar to phone entry, Volvo Sensus requires more steps. The driver has to say “navigate go to address” to select address entry. And, as opposed to simply speaking a full address, Sensus breaks the address into component parts in steps, prompting the driver for city name, street name, and street number individually. The driver is then prompted to say “finish” to initiate navigation.

That may seem like a more complicated task than MyLink, but the research discovered that address entry error in MyLink was significantly higher than in Sensus. Drivers in the study were 12 times more likely to input addresses incorrectly using MyLink than they were using Volvo Sensus. “It can be observed that only two outright failures to input the correct address occurred among participants using Sensus versus 24 failures experienced with MyLink,” the study notes.

Screen Shot 2015-03-03 at 9.15.08 AM

The MIT/IIHS study findings suggest that a properly designed and used (our emphasis) interface can reduce eyes-off-the-road time, neither the MyLink nor the Sensus system completely reduced visual demand. When compared with a traditional baseline for task duration and visual engagement — tuning a radio manually — destination address entry can be “quite extensive.”


The bottom line in the study is that evaluations that ignore “the complex intertwining of resource demands placed upon the driver paint an incomplete picture of the benefits and limitations associated with various interface design approaches and implementations.”

(Table Source: IIHS)

Craig Fitzgerald

Craig Fitzgerald

Writer, editor, lousy guitar player, dad. Content Marketing and Publication Manager at