Americans like their high-tech driver-assistance features, but more are growing uncomfortable with the concept of giving full control to a self-driving vehicle, a new Massachusetts Institute of Technology study has found.

Survey respondents reported feeling much more comfortable with technology that assists drivers, such as automatic emergency braking, rather than the idea of fully self-driving technology—and the confidence gap between the two is widening.

Drivers were asked the maximum level of automation they’d be comfortable with, and 59 percent said “features that actively help the driver, while the driver remains in control.” That’s up from 40 percent in last year’s survey.

But only 13 percent in this year’s survey said they’d be most comfortable with “features that completely relieve the driver of all control for the entire drive.” That’s down from roughly a quarter of drivers surveyed last year.

The 2017 survey—conducted by the MIT Advanced Vehicle Technology research consortium—also showed changes in sentiment by age group. This year, for drivers 25 to 34, only 20 percent said they’d be most comfortable with full autonomy; for drivers 35 to 44, it was 21 percent.

That’s a dramatic decline from just last year, when 40 percent of drivers 25 to 34 and 35 percent of drivers 35 to 44 said they’d be most comfortable handing over full control.

More on Autonomous Vehicles

Lack of trust in emerging software was a prevalent theme in responses, the survey found. Why the changes in just one year?

“The decline in confidence appears to be a multifaceted issue,” says Hillary Abraham, a research analyst at MIT AgeLab, in Massachusetts. “It seems the need for self-driving cars to work perfectly, combined with present and past experiences of low-risk technology failure ... leads consumers to believe the technology will never be good enough such that they can trust it with their lives.

“Moving forward, it will be important for automakers to address consumer concerns,” Abraham says, “potentially leveraging driver-assistance systems in such a way that consumer trust and confidence in lower-levels of automation is built up before introducing fully self-driving systems.”

Consumer Reports is part of the MIT consortium that also includes MIT AgeLab researchers, auto insurers, and various automakers. CR tests all aspects of vehicles at its track and test facility in Colchester, Conn., and provides consumers ratings and other information about vehicles and the industry, such as the development of autonomous vehicle technology.

Driver-assist features include automatic emergency braking (AEB), forward-collision warning (FCW), blind-spot warning, lane-departure warning (LDW), rear cross-traffic warning, and rearview cameras.

Drivers surveyed really liked the high-tech driver-assist features—75 percent surveyed said that they “like most” or were “very happy” with the features in their car, up from 70 percent in 2016.

When asked “How much would you consider paying for a car that completely drives itself,” 48 percent said, “I would never purchase a car that drives itself.”

Here are the reasons:

  • 37 percent: “Loss of control”
  • 29 percent: “I don’t trust it”
  • 25 percent: “It will never work perfectly”
  • 21 percent: “It’s not safe”

Of those who said they would buy a self-driving car, the largest group (27 percent) chose a price range from $25,000 to $49,999.

“Companies developing autonomous cars have a lot of work to do before consumers will be ready to turn over the driving,” says William Wallace, a policy analyst for Consumers Union, the policy and mobilization arm of Consumer Reports. “To build trust, they should share test data with the public that demonstrates safety benefits.”

The survey was conducted last month from April 10 to 30 and involved 2,976 respondents drawn from an MIT database of interested research participants and from a panel of nationwide Toyota and Lexus owners.

Most survey respondents own a car that is model 2011 or newer. Bryan Reimer, a research scientist at MIT’s AgeLab, says he expects that a generalized national sample of drivers would draw the same results.