As Google holds the first-ever driver's license for an unmanned car and Cadillac prepares its auto-pilot Super Cruise feature for production, researchers at the Massachusetts Institute of Technology remain skeptical over the technology's exciting promises.

"Knowing where the road is is only part of the answer," said Jonathan How, an astronautics professor, at MIT last Friday. "If the automated system can't handle it and throws it back to me, what am I supposed to do?"

How was speaking at an industry conference held by the university and the New England Motor Press Association, alongside panelists from BMW, Mercedes-Benz, General Motors and Kia. His attitude, as well as that of Bryan Reimer, a research engineer at the MIT AgeLab, was a sharp check against the upbeat, optimistic take held by leading automakers such as Mercedes, which is developing a system for the next S-Class that can read the entire road surface.

After MIT's autonomous Land Rover collided with a Suburban entered by Cornell University in the 2007 DARPA Urban Challenge, a government-sponsored​ competition to test fully autonomous vehicles for military use, the university has been focusing on "intent recognition," a human trait that software has not been able to mimic well.

Reimer, who studies how age and new technology affect driving behavior, said that while the initial excitement and apprehension of using an autonomous system may eventually pass, a driver's skill may also go with it.

"We need a certain amount of demand to stay focused," he said, citing a report from the Federal Aviation Administration suggesting that auto-pilot systems were reducing human flight skills. "Until driving is totally autonomous, [automation] has the potential to make the problem [of distracted driving] worse."

BMW, which since 2006 has run an autonomous Track Trainer around the Laguna Seca racetrack in Monterey, Calif., says it views the technology as a driving aid rather than a replacement. "We're trying to make a good driver better," said Tom Baloga, vice president of engineering for BMW of North America. "We're not trying to take the driver out of the loop."

While automakers cited high costs and unknown liability risks as the two biggest obstacles for autonomous systems, features such as Construction Area Assist, which would alert the driver to upcoming lane closures and work zones, and Traffic Jam Assist, automated driving activated only in heavy traffic, may reach production sooner rather than later.

"We strongly believe that this [traffic] feature can be supported with the sensing and with the algorithms," said Christian Schumacher, head of systems and technology for Continental Automotive Systems in North America.

As for price, Mercedes, unsurprisingly, was bullish about tacking on $3,000 or more for automated driving systems. "That is a lot of money for customers who are willing to pay for these kind of systems," said Sascha Simon, head of advanced product planning for Mercedes-Benz USA.

As for customer acceptance -- played out in the real world through the acceptance of Ford's MyFord Touch and other complex infotainment systems -- the learning curve is filled with even more stress and anxiety, Reimer said.

"When you saturate learning and teach people a lot of information, it can affect arousal patterns and stress before even using it," he said.