Collaborative IIHS, MIT AgeLab Research on In-Vehicle Voice Systems Released

Collaborative IIHS, MIT AgeLab Research on In-Vehicle Voice Systems Released

It’s 5:15 PM, and you’re driving on the highway as you travel home from work. You fiddle with your phone in your lap, attempting to call home while also trying to keep your eyes as much as possible on the road as you speed along. There is an accident in your lane up ahead, but you don’t know that yet. The critical seconds you spend attempting to navigate through call menus and button presses just to let your family know you’re on the way could be vital to you being able to stop in time.

If you opt for traditional manual-entry methods, how long would you keep your eyes off the road as you search for the contact link for your home phone? What about using more recently established voice-activated commands that are supposed to allow you to keep your eyes on the road while vocally instructing your device to make the call for you? In both cases, by the time you regain focus on the highway, it could be too late for you to stop for that accident coming up.

Multitasking while driving is a very present risk that most of us expose ourselves to every day – even if that multitasking is “just” thinking about something other than driving. The meteoric rise of GPS navigation and smartphones directly contributes to this potentially hazardous distracted driving. To combat this, modern cars are being designed with in-vehicle, multimodal interfaces that are intended to minimize distraction away from the primary task of driving. These systems allow drivers to input information either manually or vocally to fit their preferences. Voice-command interfaces in particular have been developed with the overt goal of allowing drivers to keep their eyes on the road and hands on the steering wheel. Some experimental studies have developed data suggesting that voice-activated systems might be less distracting overall than manual button style or touchscreen interfaces, but is this actually the case?

The Massachusetts of Technology AgeLab, New England University Transportation Center (NEUTC) and Insurance Institute for Highway Safety (IIHS) investigated multiple methods of interacting with technology inside of a vehicle, exploring the differences between two in-vehicle systems (Chevy MyLink and Volvo Sensus) and a modern smartphone (Samsung Galaxy S4). These interactions included placing phone calls both manually and vocally, and entering navigation directions vocally. In these studies, researchers found that voice interfaces decreased the amount of time drivers spent looking away from the road. However, vocal input did not eliminate visual distraction completely, as drivers often looked to the screen to confirm the voice input was interpreted correctly or to respond to information presented visually as part of the “voice” interaction.

This visual engagement associated with in-vehicle voice commands varied greatly between car models and appeared related in part to how the interface was designed. Under ideal conditions, the Chevy MyLink system demanded the least amount of interaction between driver and car as it did not require the driver to navigate through multiple menus to make a vocal selection. For example, a single voice command could be used to specify both a contact name and a specific phone (“Call Jane Smith’s mobile.”). The Volvo Sensus utilized a longer multi-step command menu system, and required drivers to verify their selections before moving on to the next portion of the menu. However, these different approaches resulted in a trade-off in accuracy, which was most evident in entering addresses into the navigation systems. MyLink users more frequently ran into difficulties getting the system to recognize what they were saying, resulting in more errors overall. Sensus users spent more time interacting visually, vocally and audibly with the interface, but experienced less errors. Excessive backtracking through technical errors may increase frustration in drivers; a behavior that is distracting in and of itself as it may draw cognitive resources away from driving and towards the problematic technology.

The AgeLab, NEUTC and IIHS researchers concluded that there were benefits and drawbacks to the voice technology in the smartphone relative to the two embedded voice systems. While the smartphone largely outperformed the vehicles’ voice interfaces in the time it took and the amount of visual engagement to enter addresses, the embedded voice systems showed advantages in less eyes off the road time to place a call and were rated by drivers as involving less work to use. Results were also mixed for manual interactions. The average time to call a number using the smartphone was shorter than when using Sensus but no different from MyLink. It should be noted that the smartphones were attached to commercially available phone stands during the study due to concerns that reaching for and trying to hold a smartphone by hand during driving may be particularly risky. It is unknown to what extent the observed demand and distraction for the smartphone condition would have differed if the phones were hand-held.

Overall, evidence is converging that voice interfaces offer a less visually demanding way to access and input information than manual alternatives. In addition, if drivers choose to engage in placing calls, the embedded vehicle voice interfaces would appear to be the most advantageous method of the ones considered in the current study. In contrast, the relative benefits of the embedded vehicle voice interfaces or the smartphone voice interfaces for destination address entry are not as clear. To read an extended summary of the research or to see the complete research reports, use the links below:

For more information on the research see the March issue of the IIHS Status report.

Mehler, B., Kidd, D., Reimer, B., Reagan, I., Dobres, J., & McCartt, A. (2015). Multi-modal assessment of on-road demand of voice and manual phone calling and voice navigation entry across two embedded vehicle systems. Arlington, VA: Insurance Institute for Highway Safety. (pdf)

Reimer, B., Mehler, B., Reagan, I., Kidd, D., & Dobres, J. (2015). Multi-Modal Demands of a Smartphone Used to Place Calls and Enter Addresses during Highway Driving Relative to Two Embedded Systems. Insurance Institute for Highway Safety, Arlington. (pdf)