NASA Finds What I Did: Level 2 Driver Assist Systems Can Make You Drowsy, Inattentive

My Experience Here

In a previous article, I discussed the Level 2 attention problem. In my testing of various manufacturers’ Level 2 driver assist systems, I’ve noticed that with extended use it’s hard to not feel a little drowsy. When I disabled auto-steer and instead used adaptive cruise control (ACC), or traffic-aware cruise control (TACC) only, I’d very quickly feel more alert and less sleepy. There’s something about the steering task that engages enough of my brain to avoid that issue.

Some readers think I’m the odd one out, because they claim the use of systems like Tesla’s Autopilot makes them feel more awake and aware, or otherwise better able to pay attention to the road. I’ve experienced that myself on short drives, but after 20-30 minutes, that effect would wear off and I’d feel like I was getting sleepy.

Past Studies

If you’re still thinking I’m lying, exaggerating, that I must be medically unusual, or that I’m secretly shorting stocks (I don’t own or short anybody’s stocks), I do need to point out that I’m not the only person to observe the drowsiness issue. A 2018 University of Utah study on this issue ran into ethics/safety concerns, because drivers were falling asleep far too often, and often weren’t aware that they had done so.

In the Utah study, one researcher’s notes of a drive read, “(Participant) is basically sleeping this entire drive. This is actually terrifying.” Research associates, fearing for their safety and the safety of other road users, complained to the university’s institutional review board. Every university and many other research institutions have these review boards to monitor studies to be sure they’re ethical, safe, and otherwise acceptable.

The review board halted the study, and the university’s Center for Distracted Driving had to redesign the study. Being more careful to exclude sleepy or intoxicated participants was enough to get the review board off the researchers’ backs, but not enough to get all of the research assistants and participants back on board.

Other studies have shown similar problems with attention. The IIHS found that once drivers get used to an L2 system, they stop paying as much attention to the road. Beyond these studies, many other aspects of the human factors of vehicle automation have been studied, and you can find a summary of many older studies here.

Long story short, yes, there’s a well established problem with drivers paying attention when using lane centering and adaptive cruise in the academic literature, regardless of anyone’s anecdotal experiences to the contrary (which I’m sure I’ll hear a lot about in the comments).

NASA Studied This Again, Found The Same Thing

NASA’s Fatigue Countermeasures Lab studies a variety of automated systems, but you’re probably wondering why NASA would study cars. With the exception of one 2008 Tesla Roadster, automobiles are not aeronautic or space vehicles, after all (NASA stands for National Aeronautics and Space Administration). The answer to that question is that NASA studies a variety of vehicles to learn how humans interact with them so that they can build or design better spacecraft.

Instead of only studying well-rested participants like the University of Utah’s redesigned study, NASA took care to have their participants have a variety of sleep amounts, just like the general population. This way, they’d know that they got a good cross section of the amounts of sleep real people get.

Participants spent almost an hour driving in a simulator, two times. This way, even sleepy drivers would be in no danger or present any danger to other motorists. If someone did fall asleep, the worst that could happen was the crash of a virtual vehicle.

On the first trip, they had full control of the simulated vehicle’s pedals and steering wheel, and controlled the virtual vehicle. On the second trip, they were only supposed to monitor an automated system like you’d find in a growing number of newer vehicles. They also had the same scenery, and “saw a flat, monotonous, two-lane road in the countryside roll by, with little traffic and no stop signs or traffic lights.”

During the drive, participants had electrodes attached to monitor brain function, and were also monitored for eye movements indicating drowsiness or sleep. After the study, participants were asked to rate their sleepiness level and to take an attention test.

The results should surprise nobody who has looked at previous studies or who have had the experiences I’ve had.

People who were supervising a vehicle scored lower on attention and higher on sleepiness than participants who were actively controlling the virtual car. Slower reaction times were also found among supervising participants compared to those actively driving. Unsurprisingly, participants who came into the study on less sleep were more heavily affected by the effect of automation, having even worse sleepiness and reaction times when supervising. However, they still did better driving than they did supervising.

Obviously, getting less sleep isn’t good for either mode (supervising or driving), but people who got less sleep generally did worse using an automated vehicle than driving themselves.

What We Should Learn From This

Obviously, the biggest takeaway is that driver assistance systems are no substitute for sleep. In theory, future fully autonomous vehicles would be a relatively safe place to take a nap, but today’s assist systems aren’t good enough yet to drive without a human keeping an eye on them. If you’re sleepy, call an Uber, pull over and nap, or otherwise don’t drive tired, regardless of how good you think your car’s ADAS is.

The other thing we should take away from this is that Level 2 ADAS systems are problematic for attention. People are better at supervising a task that they’re actively participating in on some level, and having the vehicle take too much of that driving task away leaves people vulnerable to getting bored. People who are getting too bored can fall asleep.

Fortunately, ADAS systems have been pretty good, with fairly limited injury or death considering their overall use. There have even been cases where drunk or tired motorists have been stopped by police, with the system not crashing. However, we shouldn’t necessarily assume that the Level 2 system “saved” them, as it could have contributed to them falling asleep in the first place. If someone passes out due to diabetes or some other cause that wouldn’t have been affected by the driving task, sure, Tesla and other automakers can take credit for saving a life, but if the system solves a problem it first exacerbated, there’s no credit to be had.