What will happen if we just accept that a certain number of pedestrian deaths are an inevitable part of adopting autonomous vehicles?
I’m enthusiastic about the potential for autonomous vehicles. Their great promise is that they could be safer than fallible human drivers, who kill 37,000 Americans a year. And I do believe AVs will be safer. They will not drive drunk or distracted, and they will not get overwhelmed by more information to process.
How much safer they are, though, will depend on the humans who design them and make the rules. Their driving style, whether aggressive or timid, is something that will be baked into their programming, based on real-world traffic environments and how traffic laws are enforced. The question is the same as always: Are we programming for a world that’s built for humans, or a world that’s built for cars?
The killing of a pedestrian in Tempe, Arizona, on Sunday by a self-driving Uber is showing how autonomous vehicle safety might slide from promise to nightmare.
On the evening of March 19, 49-year-old Elaine Herzberg was crossing the road in Tempe when an autonomous Uber struck and killed her. According to police reports, the car was traveling under the speed limit but showed no signs of slowing or braking for her. Uber has halted testing of all of its automated vehicles while investigations get underway, as have other companies developing AV technology. But scarcely was the blood dry before the Tempe police chief, Sylvia Moir, issued statements to the press exonerating the car.
Moir told the San Francisco Chronicle that nobody could have seen the woman “based on how she came from the shadows right into the roadway.” Moir seems unaware that autonomous vehicles use LIDAR, a laser-based system, and thus should be unfazed by shadows.
The woman was, indeed, not in a crosswalk. Bizarrely, there is a direct, curving brick path through the area, but it’s strictly ornamental: Pedestrians are forbidden from using it, and there are multiple signs posted to tell people not to use the path. The path follows what seems to be the most logical route to a nearby bus stop, and crosses the roads at narrower (and thus less harrowing) spots than the official crosswalk, which requires traversing seven lanes, counting turn lanes.
A super-weird aspect of this crash site is that it occurred at a place where a beautiful brick-paved diagonal walking path was provided across the median, along with a sign instructing people not to use it. This is beyond pedestrian-hostile design; it's damn-near entrapment. pic.twitter.com/ZaHw9bIIrR— 🚗🚌🚚🚲 (@EricPaulDennis) March 20, 2018
This is the engineering reality of much of Tempe, and much of suburban America: Designers create inhospitable environments in which to walk, then try to prohibit walking in the least inhospitable parts of those environments. And often, when someone is killed, police rush to exonerate the driver. That reflex seems to have kicked in quite quickly in Tempe. And I fear it could now push AV makers to make dangerous programming decisions.
Let’s imagine we have two autonomous vehicle companies. Call them Wundacar and Farago. They are competing for you to buy a vehicle or use their app to hail an autonomous car.
Wundacar designs its cars to tread carefully around pedestrians. When it passes one closely on the sidewalk, it slows just a bit. It yields patiently when people cross the street. It steers well away from bicyclists and other slower road users.
Farago cars do that too, but just a little less. They don’t slow quite so much. They give way to bikes a tad less. If people are crossing, they try to nose through.
Farago’s cars might get you where you are going just a few seconds faster—just enough for passengers to notice. Might you start choosing Farago over Wundacar for your go-to app? And, in the spirit of competition, might Wundacar tweak its algorithm to keep up? Just a bit. No engineer would be deliberately making it unsafe. Just… twiddle that setting there from safety factor 10 to safety factor 9.7. Let’s see if that gets people home a little bit faster. And let’s see what that does to profits.
The local police and their counterparts will help. Someone died? Well, they were off the crosswalk, so it was their fault, right? Seems it’s not a problem that the setting went to 9.7. Now it can be 9.4. Now 9. Now 8. Now 7. Now 5.
Let’s say, instead of one person, autonomous vehicles kill 10,000 Americans a year now, or even 20,000. That’s far fewer than before automation, right? And the police agree that the cars, and the companies that made them, are not really to blame. People need to be able to get where they are going quickly, right? Nobody is trying to make the roads unsafe.
And if that’s still too many fatalities, maybe the problem is too many people in the street. We don’t need those crosswalks after all. Now that we have autonomous vehicles, no senior or person with disability or teenager needs to be able to walk; they can hail an autonomous car to get to school or the doctor’s office. In fact, let’s build fences around all the roads, just to be safe.
Congratulations. You’ve criminalized moving around in the world except in a two-ton vehicle. Our traffic gets even worse. And welcome to the world of WALL-E, where people never leave their pods.
That’s not the promise of autonomous vehicles I’m still excited about today. Beyond their safety potential, I’m excited about a world where people don’t need to spend so much of their money owning a private automobile, because there’s always one that can stop by. It’ll be a shared ride so we use our roads more efficiently, don’t have to spend billions in public dollars on new highways, and can reclaim space in cities for trees and sidewalk cafes and paths for jogging and bicycling for exercise as well as transportation.
We can get there, and to their credit, many of the companies working on the technology (including Uber) have pledged to adhere to pacts like the Shared Mobility Principles for Livable Cities, which calls for things like “prioritizing people over vehicles.”
We can insist that any pedestrian death is not acceptable, just as we do for aviation, where all incidents are studied intently, and commercial aviation deaths worldwide have plummeted from 2,469 people in 55 crashes in 1972 to just 44 fatalities—and none in a passenger jet—worldwide in 2017. There have been zero deaths on U.S. airlines since 2009.
Or, we can follow the path Chief Moir is pointing toward, where we tolerate, excuse, and justify deaths on the roadways in the name of speed. Let’s let this, the first recorded pedestrian killed by an autonomous car, set a better example for what we expect of our roads, and the technologies transforming them.