When you’re driving your car, do you ever look through the rear window to make sure there’s nothing behind it in harm’s way?

Now that many cars are equipped with backup cameras, looking back is a forgotten activity.

Dr. Bryan Reimer, a research scientist at the MIT Center for Transportation and Logistics and MIT AgeLab, made the observation during a webinar focused on advanced driver assistance systems hosted by Travelers last week titled, “ADAS in the Wild.” Describing the evolution of driver attentiveness as automation is introduced to the driving experience, Reimer repeatedly made the point humans are the backup for technology today—and that they will continue to be going forward.

“The human driver is the redundancy for the automation and the support right now. When we think about advanced driver assistance systems, we’re thinking about systems where we need to be proactively managing, proactively educating,” he said, stressing a point he made several times during the course of the one-hour webinar—that educating humans about how to use ADAS features is the path forward to mitigating the “ongoing [and] undertreated health crisis on our roads.”

Reimer, who is also the founder and co-director of The Advanced Vehicle Technology Consortium, a group that includes auto makers (like BMW, Honda, Nissan), insurers (including Liberty Mutual, Travelers, Allstate and Progressive), consumer groups (including Consumer Reports, J.D. Power), the Insurance Institute for Highway Safety, among others, stressed the need for collaboration and for human-centered design of automated driving features throughout his presentation and during a question-and-answer session.

AVT seeks to understand how drivers adapt and use—or do not use—advanced vehicle technologies, explained Reimer and Joan Woodward, moderator of the session and president of Travelers Institute, the public policy division and educational arm of Travelers.

And they’ll be using and adapting to assistive technology for many years to come, according to Reimer.

“Level 5, highly automated vehicles that work anywhere a human would drive, will never exist,” Reimer told Woodward when she asked about the timeline for fully driverless cars coming onto the nation’s roadways. “We will never develop an automated system to go out in a blizzard to go get ice cream,” he said.

While Level 4 vehicles are being tested on roads today (in Arizona, Texas and San Francisco), these systems are not perfect, Reimer said, going on to predict that “over the next century, highly automated mobility and lots of Level 4 systems will exist moving us from point A to point B seamlessly.” (Editor’s Note: The National Highway Transportation Safety Administration defines a Level 4 system as one that is fully responsible for driving tasks “within limited service areas” while occupants act only as passengers and do not need to be engaged.)

But “for the foreseeable future, these are evolving technologies and we’re still driving,” he stressed.

Reimer advocates moving away from the “levels of automation” of vehicle technology, noting that it has created confusion about what the driver’s role and responsibility as technology evolves. His preference is for simplified terminology: “I am driving and I’m responsible for all the aspects of vehicle control and safe mobility. Or I am riding, much like the yesteryears of being a passenger” in a taxi.

“Consumers have a difficult time differentiating between more than a binary A-B decision….I’m driving now, but I also need to monitor the automation and perhaps maybe I’m not quite fully responsible. You can see the legal questions of this developing,” he said.

Reimer also reported that drivers are happy to continue to have control of their vehicles. He displayed survey data collected through a partnership with J.D. Power, showing that “drivers are looking for automation that helps assist them. They’re not looking to be replaced. They’re not looking for just automation that works in emergency situations only,” he said.

In fact, the data he displayed suggested that the appeal of full self-driving has diminished over the years. It showed that while 23.7 percent of respondents said they would be comfortable with full self-driving in 2016, only 12.8 percent expressed that comfort five years later in 2021. In contrast, the option “driver assist,” defined as “features that actively help the driver [who] remains in control,” was selected by roughly 40 percent of drivers surveyed in 2016, rising to 54 percent in 2021.

Does It Work?

Reimer and Amanda Mezerewski, vice president for the personal auto insurance product at Travelers, provided slightly different takes on the benefits of ADAS so far, with Reimer suggesting that assistance features haven’t quite lived up to the benefits that insurers and manufacturers envision when they imagine the future of automation, and Mezerewski noting that takeover features have been more beneficial than alerts and nudges. Both agreed that the driver—and driver attentiveness—are the most important factor in making roads safer.

“Lane departure warning systems were hailed as a major safety breakthrough and we found that they did not work to the degree we expected….Even the benefits of automatic emergency braking, which have led to collaborative mandate across the industry, really are not showing the real-world benefits that we once perceived should exist,” Reimer said, offering the following reasons for the gap:

• Often drivers don’t have the skills and understanding necessarily to work with these technologies and successfully leverage them in the real world. “Education is a need.”
• Many systems require driver management and oversight.
• Technologists often assume ideal performance from both the human and the system, “yet we all seem to too frequently get diverted from the roadway to answer a phone call or send a text.”
• Infrastructure is often less than ideal. “Three foot craters in the road during New England winters don’t marry with high technology easily,” he said.

“What’s key here is that the understandings of these systems are complex. How drivers in the wild, engage, and where they use—and where they do not use—technologies is more than one industry alone can solve or one company can focus on,” he said, calling for cross-industry collaborations like the AVT, and also for private-public partnerships.

Frequency Drops Offset Severity Jumps

Mezerewski showed slides compiling the results of studies by the Highway Loss Data Institute (IIHS-HLDI), which illustrate the reduction in insurance claims frequencies associated with various driver assistance features.

Related article: Why ADAS Drives Claims Severity Higher (It May Not Be What You Think)

“When we look at technologies that warn the driver versus similar technologies that act on behalf of the driver, the implementation that acts on behalf of the driver is typically associated with larger reductions,” she said, summarizing a series of bar graphs on which bars represented changes in claims frequency by coverage. The downward bars for property damage and bodily injury claims were much larger for front auto brake systems than for forward collision warnings for example.

“When we look at systems designed to help people park their cars, the frequency reductions associated with parking sensors and rear camera are more modest as compared to rear automatic emergency braking, which will automatically apply the brakes if something is detected in its path moving forward,” she added, giving another example.

Recognizing that certain ADAS features can only be purchased as part of a bundle, HLDI has also analyzed the claims frequency drops associated with these packages, finding that some approach 40 percent for property damage liability and bodily injury liability coverages.

Turning to claims severity, the bars on the graphs for collision and property damage liability claims costs go up instead of down, confirming that ADAS features contribute to a higher average repair costs when crashes happen. Mezerewski explained that the cost of the technology is one reason for increasing severity. In addition, because the technologies “are very good at eliminating low dollar claims—think really minor bumps and crashes—this results in an overall increase in the average value for the claims that do remain,” she said. (The concept of “mean shifting” is further explained in an earlier Carrier Management article, “Why ADAS Drives Claims Severity Higher (It May Not Be What You Think)”)

Even with the severity jumps, however, Mezerewski noted that the frequency declines “far outweigh any increase in average costs.” In other words, “the impact that ADAS bundles have on the reduction of total [property] damage loss dollars is still substantial,” she said.

At one point, she noted that while “the cost of repairs when a collision does occur is absolutely higher,” Travelers gives a discount for drivers that that AEB on their cars in recognition of the fact that frequency drops push down aggregate loss costs.

She noted that it’s going to be a while before insurers notice the aggregate loss cost benefits of ADAS impacting their books of business, however. “It typically takes decades after a feature is introduced before [it] is installed on most vehicles on the road,” the Travelers leader said, explaining that new features are typically placed in higher-end luxury vehicles first, taking “five to 10 years to become optional on new model years,” subsequently working their way registered fleets over time.

Today, the feature that’s on the most vehicles is the rear camera, with a 2018 federal mandate giving them a boost. Over half of registered vehicles had them in 2022, Mezerewski reported, again referencing IIHS-HLDI data.

Front AEB, the technology that IIHS-HLDI and member companies have been advocating for the longest time, is also becoming more common, partly as the result of a voluntary commitment made by 20 auto manufacturers to make the technology standard back in September 2022. By 2027, HLDI estimates nearly half the vehicles out on the road with this feature, Mezerewski reported.

Education and Design Matter

During his presentation, Reimer talked about the time it would take to educate drivers in the use of ADAS features, and about the need to make driver-friendly design changes.

“As we move forward,infusing what automation has to offer with [actions] of an attentive driver, we should see severity and frequency of accidents reduced. But unfortunately I think for the foreseeable future, as we continue a bumpy on road of more assistance, we’re probably going to see the frequency of collisions and excessive losses increase,” he said, going on to described the levels of inattentiveness common among drivers today.

“At the moment, driver skills, situational awareness and an understanding of the driving environment are continuing to decrease,” he said. “Automation leaves drivers perceiving that they need [to pay] less attention to the road.”

Reimer views driver monitoring and support systems as important keys to the promise of road safety. “Driver attention may be at historic lows, and there’s a clear indication that assistive automation may exacerbate this problem,” he said, endorsing the idea of putting “same fundamental elements of artificial intelligence” that are now focused on “sensing and path planning technology to drive the vehicle itself,” and instead refocusing the efforts on “monitoring and connecting the driver to these systems.”

“Instead of racing to automate, we still need to produce and enhance systems that include humans,” he said. “Given the already high loss ratios, we can see that the industry alone is not going to solve this. Government needs to take an active role, with everybody willing to collaborate…

“How drivers in the wild, engage, and where they use and where they do not use technologies is [a problem] more than one industry alone can solve or one company can focus on,” he said.

During the Q&A part of the session, when Woodward asked Reimer the most surprising finding of the research he’s done, Reimer responded that “design matters a lot. Marketing matters an enormous amount, but design matters. We see similar conceptual features on the road influencing driver behaviors quite differently,” he said, adding that because features are early in their deployment it would like take a “good part of the decade” to understand what works well to support safety and mitigate risk.

Asked to identify the biggest misconception about ADAS technology today, he said it’s the idea “that vehicles are self-driving,” and that ADAS systems can be beneficial without greater driver understanding. “The literature around automation really tells us that with increased automation comes increased needs for driver education…We need to understand things such as where can I use it? What does it do for me? What do these alerts mean?”

But MIT’s research reveals that when drivers hear alerts, “they’re going to the instrument cluster or the center stack of the vehicle searching for information on what’s wrong. They don’t know. They’re looking to understand. And education is very, very difficult because nobody owns the education problem,” he said. “Manufacturers don’t touch consumers. Dealers are the ones that are front lines, but they don’t have the expertise and the understanding of all the different vehicle systems out there,” he said, noting that the likes of Consumer Reports and AAA are left to do the education, but they’re “talking about lots of different flavors of ice cream” rather than one common system.

Asked whether there is any data showing that assistance technologies actually erode driving skills, Mezerewski said, “The distraction is real.”

“We would continue to reinforce [that] the important safety feature continues to be the driver. There is some data on the causation of these technologies diminishing driver skills. Overreliance is really the problem there—over-relying on the technology and then drifting,” she said.

Earlier, during his opening presentation, Reimer referenced a real-world study of GM’s Super Cruise system by a colleague (Pnina Gershen), showing that when disengaged drivers with both hands off the steering wheel were asked to take control, it took them 33 percent longer than drivers with one hand on the wheel (2.4 seconds).

Responding directly to the question about declining skills, Reimer noted that in the aviation world, “the FAA has been very aware for over a decade now that automation is eroding pilot skill.” He added, “I think they’re going to see the exact same thing with advanced driver assistance systems, partial automation and at some point features such as highway pilot that get us from point A to point B on the highway.”

Referring to earlier forms of vehicle automation, power steering and automatic transmissions—he asked, “How many of us can drive standards these days,” referring to stick shifts or standard transmissions. Insurers and drivers need to accept the fact that skill level does change, and think about “how do we begin to build that and what the minimum acceptable skill really needs to be.”

“This concept of getting your driver’s license at 16, 17 years old and never really getting any education after that is not something that’s going to develop safe systems long term,” he concluded.

“There can even be a question on how do we define ‘inexperienced,'” said Mezerewski. Instead of the typical meaning put a younger person licensed for a year in that classification, inexperienced could mean someone who’s driving a vehicle with all these technologies for the first time, she suggested. “I think that’s also a really important implication as we think about the data over time.”