This has been known for a long time. We now have evidence that human drivers pay less attention to what is happening on the roads after the Autopilot self driving tech has been enabled on Tesla cars.
The study reveals the awkward in-between phase we are in. While self-driving technology has improved enough to handle many aspects, it can't take over all the driving.
This is more dangerous than fully automated driving. People assume that they don't need to pay attention to all aspects of driving.
The researchers published a paper titled "Visual behavior patterns alter before and after Autopilot Disengagement." "Before disengagement drivers focused less on the road and more on areas that are not related to driving, compared to when they switched to manual driving."
"The lower proportion of off-road glances prior to disengagement from manual driving was not compensated for by longer glances ahead."
Autopilot is not capable of driving a car in all situations, despite being extremely capable. Tesla claims that Autopilot was "designed to help you with the most burdensome aspects of driving" but that it still requires active driver supervision so that the vehicle can drive itself.
Researchers from Massachusetts Institute of Technology (MIT), conducted an ongoing study about driving and advanced technology. They examined driver postures and facial positions to determine where their eyes were focused. The team used data from 2016 to compare 290 instances of Autopilot disengagements and compared their behavior after.
The study used data from nearly 500,000 miles (over 800,000.km) of travel.
The vast majority of off-road glances that were recorded when Autopilot was activated, were focused on the large screen in the middle of every Tesla car's dashboard. Researchers found that Autopilot was activated in 22 percent of the glances, while only 4 percent were with Autopilot off.
Autopilot enabled, so off-road glances took longer. Manual driving made it more common to glance at the side windows, side mirrors and rearview mirrors. Researchers also created a simulation model that could estimate glance behavior using a larger number of data points.
According to MIT researchers, "This behavior change could be due to a misunderstanding of the system's capabilities and limitations. This is reinforced when automation performs relatively poorly."
Autopilot, which is an autonomous system like Autopilot, should monitor drivers and adjust their behavior according to how attentive they are. Autopilot currently uses pressure to determine if someone is paying attention.
Researchers also believe it's important for drivers to be fully informed about what self-driving technology can and cannot do. Tesla informs drivers that attention is still needed, but it calls its latest software update Full Self Driving.
This latest study does not show a link between safety and attention span. Therefore, there is no conclusion about whether Autopilot is safer than manual driving. It makes drivers pay less attention on the road, which is what is clear.
The researchers conclude that the model in this instance can be used to perform safety benefit analyses through simulation, which can help inform policy-making and design of driver support systems.
The research was published in Accident Analysis and Prevention.
