To trust or not to trust? Autonomous vehicles are the future.
Evan Hoye stares at three large computer screens that display a simulated road, complete with lanes, roadside fences and fields. He views this scene from a driver’s seat set-up, which is furnished with a steering wheel, accelerator and brake pedal.
He “starts” the vehicle. A motor sounds. His foot hits the accelerator. Suddenly, he’s watching the simulated lane marker and roadside scenes fly by. A few seconds later, Hoye, 24, of Ann Arbor, presses a little button on the “dashboard.”
Then he drops his hands from the steering wheel, takes his eyes off the “road” and begins to play a simple game on a tablet to his right.
Hoye is officially “riding” in an autonomous vehicle (AV). He is sampling one of two tests for studies on human interaction with AVs, or “driverless cars.” The studies were designed by Lionel Robert, associate professor of information in the University of Michigan School of Information (UMSI).
The tests simulate a variety of scenarios people might encounter in an AV. For example, moments after Hoye drops his hands from the steering wheel, the image of a car appears in the next lane. A voice announces, “Stopped vehicle ahead. No action needed.” Hoye keeps poking the tablet screen. Soon after, the voice repeats the same statement about a similar image. Hoye minds the tablet.
Then comes the command: “Stopped vehicle ahead. Action needed.” Evan quickly switches from tablet to steering wheel. He “drives” the “car” around the image of a stopped vehicle right in his lane.
Such simulations will soon be reality. AVs are on the way to everyone’s driveway, like it or not. And in fact, many don’t like it, at least not yet. For them, it will be a matter of time – and of trust.
“In American culture, driving has always been about independence; it’s like a rite of passage,” says Robert. “You become master of your own destiny. And all of a sudden, we’re asking people to turn that over to a vehicle.”
The first of Robert’s two AV trust studies was commissioned by U.S. Army personnel, who wanted to reduce the number of troops needed for various tasks. AVs were one possible way to do that. If there were two soldiers in one vehicle, an AV could take over the task of driving and require just one soldier, who could perform other tasks, such as logistics, watching the enemy, etc.
At first, Robert says, “drivers didn’t trust the car’s autonomy. They didn’t feel comfortable letting it drive…. They were less focused on the other tasks and more worried about what the vehicle was doing.”
When the AV informed the soldier of its current status, the soldier trusted the vehicle much more, and was able to perform other tasks better, Robert says.
The second test, “Trust, Control and Risk in Autonomous Vehicles,” involving civilians, focused on expectations and “the impact, and timing, of explanations” given by the vehicle for what it is doing, Robert says.
“For us as human beings, trust is built on expectations,” Robert says. “I expect you to do something. If you do it, I trust you. If you don’t do it, I lose trust in you.”
This test, which is the one Hoye navigated, offered three choices: The driver does nothing; the AV provides a status update; and the AV provides a status update and tells the driver to take control.
“We thought the third condition would be preferred, but that wasn’t always the case,” says Robert. “Some people preferred just to get the status update and make their own decisions about when they would take over the vehicle.”
Hoye’s response reflected some of that desire for some kind of control. When the test screen at one point showed the car swerving for instance, he said, “I felt more comfortable allowing that to happen because I could also make the choice to intervene.”
More tests are coming, and only time will tell if more trust will follow. Robert believes the complete changeover to AVs may take as long as 20 years or more – not unlike the transition from the horse-and-carriage age to the horseless carriage age. It’ll be interesting, he says, and a little messy.
“You can imagine a world of smart cities with AVs talking to each other, interacting with safe, efficient, effective driving,” he says. “But meanwhile, you’re going to have AVs next to Mom and Pop. And they’re not going to drive the way we would want them to drive. So there are always going to be limitations about how safe and efficient AVs can be as long as they’re riding next to a human driver.”
Nevertheless, he adds, “I think we’re living with the last generation of people who are going to make a living driving anything.”
View a three-minute video on autonomous vehicle simulations in the U-M MCity lab.