A self-driving vehicle is speeding down a city street when a garbage truck turns across its path. Does the car’s onboard computer decide to hit the back of the truck? Or swerve into a nearby crosswalk, minimizing risk to its own passengers but almost certainly harming innocent bystanders?
It’s a version of the trolley problem, the philosophy-class ethics conundrum made real by the advance of autonomous vehicle technology in Detroit and Silicon Valley. Should regulators permit automakers to value the safety of passengers over pedestrians? In the year since a computer-driven Uber hit and killed a woman near Phoenix, researchers and journalists have pointed out ways that A.I. might be more focused on the passenger experience than the people nearby. (Autonomous vehicles may have problems seeing people of color, for example.) Developers of vehicular intelligence are already arguing that it’s the pedestrians who must be reprogrammed, as they were when the first automobiles hit the streets at the turn of the 20th century. How will regulators evaluate the ethical trade-offs built into the code?
A clue can be found in a real-life iteration of the trolley problem that has been quietly playing out for the past 10 years. American consumers have increasingly opted for cars that make drivers safer and pedestrians more vulnerable, as pedestrian deaths have reached crisis levels. More than 65 percent of new vehicles sold in the United States are pickups or SUVS, up from 49 percent a decade ago. (Ford has all but stopped U.S. production of cars.) The SUV revolution, a Detroit Free Press/USA Today investigation concluded last summer, is “a leading cause of escalating pedestrian deaths nationwide.”
Last month, the National Highway Traffic Safety Administration released projections for roadway fatalities in 2018: For the second year in a row, overall traffic deaths fell, while pedestrians fatalities rose, and cyclist deaths jumped by 10 percent over 2017. “America’s roads are safe increasingly for only those who drive on them,” Aaron Short wrote at Streetsblog.
It caps a bloody decade on the road for anyone outside a car. In February, the Governors Highway Safety Association estimated that more than 6,200 pedestrians died in 2018—the highest number in a generation, and an increase of 50 percent from 2009. The GHSA says SUVs are partially responsible: Their involvement in fatal crashes is up 50 percent, compared with 30 percent for sedans. (If you’re wondering if lives were saved overall by this shift, thanks to the prisoner’s-dilemma incentive to adopt tanklike vehicles, the answer is no: The fatality rate per 100 million vehicle miles traveled is the same as it was in 2009—it’s just that pedestrians account for 16 percent of traffic deaths now, up from 12 percent.)
The Detroit Free Press investigation showed that NHTSA knew four years ago that the proliferation of pickups and SUVs was putting pedestrians at risk thanks to heavier vehicle weights, higher bumpers, and compromised visibility. “Pedestrians are 2-3 times more likely to suffer a fatality when struck by an SUV or pickup truck than when struck by a passenger car,” the agency concluded in 2015. Children between 5 and 9 years old had a fatality risk four times greater in collisions with light trucks and SUVs than with cars. The regulators did nothing. If another product saw its nonuser death count spike by 50 percent in 10 years, consumers would revolt and Congress would make a big show of getting to the bottom of it. Automobiles are different.
Clearly, SUVs are not entirely to blame for these deaths: Driving miles have ticked up since the recession, every driver now has a smartphone in hand, and car-centric infrastructure has not kept pace with interest in walking and biking. Joe Cortright at City Observatory makes the comparison with Europe, where lower oil prices and economic recovery have also coincided with ubiquitous smartphone use and big market gains for SUVs. But between 2007 and 2016, he notes, European pedestrian deaths are down 36 percent. In 2007, pedestrian deaths per 100 million people were higher in Europe than in the U.S. (which stands to reason, since Europeans walk much, much more than we do). Now, those numbers are reversed—pedestrian deaths per 100 million are nearly twice as high in the U.S.
It’s not clear exactly why the United States and Europe are diverging so rapidly on pedestrian deaths. But Europe’s approach to new cars is different: The Euro NCAP (New Car Assessment Programme) that supplies safety ratings for European cars pays attention to how new vehicles affect the people around them. “Vulnerable Road Users” is one of four characteristics on which vehicle safety is addressed. Head impact, pelvis impact, and leg impact for a pedestrian in a collision are evaluated. A color-coded grid of impact ratings is superimposed on a projection of the car front. Automatic systems are assessed: Of the Kia Ceed, for example, the NCAP notes, “The AEB [automatic electronic braking] system fitted as standard responds only to other vehicles and does not detect vulnerable road users like pedestrians and cyclists.”
In America, by contrast, NHTSA ratings have three primary categories: frontal crash, side crash, and rollover. What happens to the people you hit? Who knows.
If that’s the logic that carries forward into regulatory scrutiny of autonomous driving technology, we’re going to see a lot more incidents like the 2018 Phoenix crash, in which an Uber Volvo SUV driven by a computer hit and killed a woman with a bicycle. American regulators already faced one trolley problem and came to the conclusion: Who cares?
