In response to German officials asking Tesla to stop using the term AutoPilot for its driver assistance system, the automaker commissioned a survey of owners to gauge what they really understood. The good news from the results is that the glass is 93% full. The bad news for Tesla and safety is that it’s a very large glass and even 7% misunderstanding is an awful lot of drivers.
When asked “Has the name “Autopilot” caused you to believe that the car is fully autonomous, meaning that it does not require the driver to be supervising the car?” 7% of the German Tesla owners surveyed responded yes. Let’s put that in some context. German research firm puls Marktforschung GmbH surveyed 675 Tesla owners in Germany.
Click here for the Tesla AutoPilot Awareness Survey
Given the extremely lax requirements to get a driver’s license in the U.S. relative to what it takes in most of Europe, it’s probably fair to say that most German drivers probably understand more about driving and the cars they are operating than Americans. However, for the sake of argument, let’s assume that a similar proportion of American Tesla owners understand the capabilities of AutoPilot. Now let’s extrapolate that the total population of American drivers. According to the National Highway Traffic Safety Administration (NHTSA) there were 214 million licensed drivers in the U.S. in 2014. Even a mere 7% of that population would be 15 million drivers.
Given the cost of current Tesla products, it’s customer base is considerably more affluent and generally better educated than the overall average of the population. Thus, In all likelihood, the Tesla owner base probably has a better than average understanding of the limits of AutoPilot so the numbers that would think it is capable of full autonomy would be much higher but let’s go with that 14 million number.
I’m all for improving safety on the world’s roads. With more than 1 million fatalities around the world every year, the human cost is just too high. I do believe that in time, autonomous vehicles will be able to dramatically cut that number while also reducing energy consumption and congestion.
But context is also important. In 2014, there were more than 3 trillion vehicle miles traveled in the U.S. and 4.4 trillion passenger miles traveled. Out of all those miles, there were 32.675 people killed including almost 5,000 pedestrians and another 1,000 cyclists and others. While it’s widely estimated that between 90 and 95% of crashes are caused by human error, that does not in any way imply that 90% or more of drivers make errors that lead to crashes.
In fact, in 2014, there were a little more than 6 million total reported crashes and even if we assume that no drivers crashed more than once, that only represents 2.8% of all drivers in the country. More than two and a half times that percentage think AutoPilot is capable of fully autonomous driving.
I’m glad that Tesla is pushing technology forward. With their efforts, the rest of the industry might not be moving nearly as fast to electrify vehicles. Tesla has also made safety a major priority as well as automotive cybersecurity.
However, having spent some time on the public relations side of this business I’m also fully aware that the job of the communications staff is to present their company and its products in the most positive light. It’s also the job of the rest of us to look beyond the information that the PR staff choose to highlight in order to get a more complete picture of what is actually happening.
I’ve driven the Model S with AutoPilot and its remarkably capable. My friend Alex Roy at TheDrive.com has accomplished some amazing feats with AutoPilot equipped Teslas over the past year. But Alex is far from a typical driver and I believe it does a disservice to represent statistics about vehicle automation in this way and I still believe that some of AutoPilot’s capabilities do not sufficiently factor in the human element.
Dr. Bryan Reimer of the Massachusetts Institute of Technology heads up the Advanced Vehicle Technology (AVT) Consortium. Dr. Reimer and his team have been conducting a study over the past year with drivers in a number of different vehicles from Tesla, Volvo, Land Rover and Mercedes-Benz with a range of technologies. Among the questions they are researching are how driver’s adapt to technologies, how does understanding/acceptance over time change, which systems are used and how does behavior change with that use.
While the study is still ongoing, Reimer has presented some initial observations at conferences including the recent TU-Automotive ADAS to Autonomous Vehicles conference in Detroit. In short, system performance still remains varies widely, the human machine interface is far from optimal and driver education about these systems needs a lot more work.
Control hand-off between the automation and manual operation is also one of the biggest issues. In fact this may be at the heart of some of the reported Tesla accidents this year where driver’s claimed they were using AutoPilot while Tesla claims the system was not active according to its telemetry data. What Tesla has not reported in its responses was exactly when the system deactivated. If the drivers were not aware of the deactivation, it’s possible they thought the system was still operational when it was not.
When a glass this size is this full, even a small spill can make a big mess. If we push intermediate technologies like what we have today too hard, we risk additional crashes while also damaging longer term consumer acceptance of the technology when it’s more mature. Automakers and suppliers need to focus more on the areas that Dr. Reimer and his team are studying at MIT and less on trying to make the case for selling the technology available today.
The author is a senior analyst on the Transportation Efficiencies team at Navigant Research and co-host of Wheel Bearings
