Drivers fidget with electronics and take both hands off the wheel more often as they develop trust in automated systems, new research from the Insurance Institute for Highway Safety and the Massachusetts Institute of Technology’s AgeLab shows.
To investigate how experience with automation affects driver disengagement, the researchers studied the driving behavior of 20 Massachusetts-based volunteers over a month’s time as they gained familiarity with advanced driver assistance features, examining how often they removed both hands from the steering wheel or took their attention away from the road to do things like use their cell phone or adjust the controls on the vehicle’s console.
One group of 10 drove a Land Rover Range Rover Evoque equipped with adaptive cruise control (ACC), which automatically keeps the vehicle traveling at a speed chosen by the driver while maintaining a pre-established following distance. Another 10 drove a Volvo S90 with both ACC and Pilot Assist, a partially automated system that combines ACC with lane-centering technology that keeps the vehicle positioned laterally in the travel lane.
Under the classification system developed by SAE International, the levels of automation range from 0 (no automation) to 5 (fully self-driving). Level 1 systems can assist the driver with one driving task; ACC fits into this category. Level 2 systems, such as Pilot Assist, can assist with two tasks. Level 2 is the highest level of automation available in production vehicles today.
When the drivers first received the vehicles, there was little or no difference in how frequently they showed signs of disengagement, whether they were driving manually, using ACC or using Pilot Assist. After a month, however, they were substantially more likely to let their focus slip or take their hands off the wheel when using automation, and the impact of Volvo’s Level 2 system was more dramatic than that of ACC alone, says IIHS Senior Research Scientist Ian Reagan, the lead author of the study.
“Drivers were more than twice as likely to show signs of disengagement after a month of using Pilot Assist compared with the beginning of the study,” Reagan says. “Compared with driving manually, they were more than 12 times as likely to take both hands off the wheel after they’d gotten used to how the lane centering worked.”
Pilot Assist and similar systems like Tesla’s Autopilot, Cadillac’s Super Cruise and Mercedes-Benz’s Intelligent Drive are not designed to replace the driver. They have trouble negotiating many common road features, so the driver must be in control at all times. However, with the automation managing steering and speed — quite well in some cases — it’s easy for the driver to lose focus.
“This study supports our call for more robust ways of ensuring the driver is looking at the road and ready to take the wheel when using Level 2 systems,” says Reagan. “It shows some drivers may be getting lulled into a false sense of security over time.”
Earlier this year, IIHS issued a series of recommendations for improving how such systems monitor whether the driver is paying attention and how the systems react when that focus falters (see “Automated systems need stronger safeguards to keep drivers focused on the road,” March 12, 2020). The European New Car Assessment Program recently launched ratings for driver assistance systems that assess those capabilities in addition to how well the systems control the vehicle’s speed and steering. However, U.S. regulators have yet to develop similar ratings or standards for evaluating partial automation systems.
When it comes to the effect of technology on driver behavior, the new study illustrates some clear differences between stand-alone ACC and the combination of ACC and lane centering.
In contrast with the disengagement observed with Pilot Assist, drivers of the S90 were less likely to become disengaged while using ACC than while driving manually, and that didn’t change much over time. However, with the availability of Pilot Assist, only 4 out of 10 drivers used ACC alone after gaining familiarity with the systems, and they used it infrequently.
Drivers of the Evoque, who used ACC often, were more likely to look at or pick up a cell phone while using the assistance technology than when driving manually, and that tendency increased substantially as they grew familiar with ACC. On the other hand, increased familiarity did not result in more frequent texting or other kinds of cell phone manipulation known to increase crash risk (see “Large naturalistic study provides new window on driver distraction,” Dec. 10, 2015). Unlike drivers using Pilot Assist, drivers using ACC in both the Evoque and the S90 weren’t any more likely to remove both hands from the wheel than when driving manually.
These differences could be important in weighing the impact on crash risk from ACC versus partial automation that combines ACC with lane centering. Field tests have suggested that ACC, by controlling speed and following distance, may have safety benefits that go beyond those provided by forward collision warning and automatic emergency braking. But neither field test data nor recent analysis of insurance claims shows similar safety benefits with the addition of lane centering (see “Benefits from advanced driver assistance systems are growing, new HLDI study finds,” April 16, 2020).
At the same time, a number of high-profile crashes involving partial automation have demonstrated how dangerous too much trust in technology can be.
“Crash investigators have identified driver disengagement as a major factor in every probe of fatal crashes involving partial automation we’ve seen,” says Reagan.