Updated Nov 21, 2018 - Technology

Why driverless cars could save far fewer lives than expected

Illustration of a crash dummy head with a band aide on its temple

Illustration: Aïda Amer/Axios

Autonomous vehicles have the potential to save lives — but they're unlikely to prevent as many deaths as we've been led to believe.

The big picture: Some automakers and politicians have suggested that autonomous vehicles will sharply reduce deaths on the road, if not eliminate them. The reality is that uneven deployment and technological limits mean AVs may save far fewer lives than the hype implies.

By the numbers: 37,133 people died in motor vehicle crashes in 2017. Government agencies like the National Highway Traffic Safety Administration and auto companies ascribe 94% of serious crashes to human error. Thus, the logic goes, take the human out of the equation and most of the crashes disappear.

  • GM, for example, holds out the promise of a "zero crashes" goal: "Our Cruise AV has the potential to provide a level of safety far beyond the capabilities of humans. As our experience and iterative improvements continue, we will advance closer to our zero crashes vision.”
  • And Sen. John Thune (R-S.D.) wrote last month: "In 94 percent of these cases, human error is a factor. One day, self-driving vehicle technology should help drastically reduce this tragic loss of life."

Reality check:

  • That often-quoted 94% causation rate is derived from a 2008 government study that was never intended to be applied to autonomous vehicles.
  • Driver miscalculations and misbehavior cause most crashes.
  • Three leading causes of fatalities — drunk driving, not wearing a seat belt and speeding — won't necessarily go away with the removal of the driver. A hurried driver might disengage the law-abiding technology if he's late for work, for example.
  • And, AVs aren't yet better than humans at perceiving their environment.

A different take: The Casualty Actuarial Society has created an Automated Vehicles Task Force to research the technology’s risks and its implications for insurance and risk management.

  • It found that 49% of crashes contain at least one limiting factor that could disable AV technology or reduce its effectiveness.
  • Autonomous vehicles are likely to present different risks and their own failures, they point out.

And there are even more reasons AVs won't solve all fatalities on the road.

  • About 51% of crash fatalities occurred on rural roads, where economics — and technological limits — don't support the deployment of AVs anytime soon.
  • Instead, AVs are likely to appear first as part of low-speed, city-centric ride-sharing services.

Factoring all of that in, Philip Koopman, a professor at Carnegie Mellon University and founder of Edge Case Research, estimates the number of deaths prevented by AVs would be around 10,000 a year — only about 1/4 of the current total.

"They can expect to save lives, but not nearly as many as the hype would have you believe."
— Philip Koopman

The introduction of advanced driver assistance systems like automatic emergency braking and lane-departure warning systems are already helping to save lives. And because they're more widely deployed, they may be nearly as effective in lowering the death rate, Koopman says.

The bottom line: The safety of autonomous vehicles can't be determined by today’s standards — because the factors that cause crashes today may or may not cause them in an automated vehicle era.

Go deeper:

Go deeper