AP Photo/Eric Risberg

Announcing a $4 billion budget proposal to support driverless vehicles, Transportation Secretary Anthony Foxx said earlier this month that the technology had “enormous potential to save lives, reduce greenhouse gas emissions, and transform mobility for the American people.”

It’s not just the Obama administration that has touted the potential environmental benefits of driverless cars: Automakers have even sought to use the technology to count towards fuel-economy standards.

But not everyone is so sure that a future of robot cars will be such a good thing for the environment. Some environmentalists are now warning that driverless cars, which could start hitting the roads within the decade, may actually be a net loss for the climate.

“There’s a utopian vision of what this looks like, but there’s also a dystopian vision,” said Roland Hwang, director of the Natural Resources Defense Council’s transportation program. “There are clearly a lot of benefits associated with this technology, especially if we harness it and push it in the right direction … But we could create a situations where we might undermine our clean transportation systems.”

The utopian vision looks something like this: Cars that can connect to each other and infrastructure like traffic lights could platoon with each other and plot out the most efficient routes, making every car trip cleaner than it would be with a human-driven car. And a shared pool of autonomous cars—think Uber without a driver behind the wheel—makes it possible to travel around a city without ever owning, or even driving, a car. Plus, cars that are programmed not to crash can be lighter, making them more efficient.

But experts are projecting that the newfound efficiency could create demand, which could mean more cars on the road and more miles traveled in them—and all the emissions that come with them. 

Connected cars can squeeze more vehicles onto existing roads (connected by computers, there’s no need to leave space between them to account for human reaction time for braking or swerving), cutting back on traffic jams and making commutes faster. That creates a concept known as induced demand, similar to when a lane opens up on a freeway and then fills up with cars.

And once fully autonomous cars take human drivers out of the equation, a commute suddenly becomes a chance to get a jump start on work, or catch up on sleep. Don MacKenzie, an assistant professor at the University of Washington, said that makes driving a more attractive option.

“Driving time is a cost imposed on you as a driver,” said MacKenzie, who said the cost of time spent behind the wheel can be considered a negative factor of car ownership, like insurance or a car payment. “With this technology, that cost won’t go to zero, but it’s cut significantly … [and] you can stimulate a lot more potential travel.”

MacKenzie is the coauthor of a study that projected as much as a 60 percent increase in vehicle miles traveled with automated technology. Automation, his study found, had the power to “plausibly reduce road transport greenhouse-gas emissions and energy use by nearly half—or nearly double them,” depending on how things played out.

Add on top of that the entire populations of people who aren’t driving now, but could with the help of a computer—like the elderly or the disabled—and the roads suddenly look a lot more crowded. Getting cars off the road is the surest way to reduce their emissions, MacKenzie said, so it’s possible that an increase in driving “could basically swamp the improvements in energy efficiency that we can get.”

A study by the Atlanta Regional Commission predicted that in the year 2040, the number of daily trips and the average trip length will increase due to the better commuting experience. The commission also predicted a decrease in public-transit trips by as much as 42 percent over the baseline in the most extreme projection, which projected total market penetration of autonomous vehicles and myriad benefits, such as reduction in parking cost and lower travel time.

“With the flexibility of shared rides, instead of large buses and light rail, those trips could be done in a smaller, on-demand or shared arrangement,” said Guy Rousseau, a traffic modeling manager who worked on the projections. “When you provide greater mobility throughout the region, then clearly it has an impact on transit. There’s still a demand for public transit, but it goes down.”

Then there’s the potential for empty cars on the roads. Driverless taxis zipping directly from customer to customer are great, but a car that, say, circles the block when there’s no convenient parking space is simply spewing emissions. Dan Fagnant, an assistant professor of civil and environmental engineering at the University of Utah, said that the ease of having a car that meets you where you are could mean cars spend more time unoccupied.

“Let’s say my household has one vehicle rather than two,” said Fagnant. “I take the car from home to work, and then I send it back home where it picks my wife up. She might go stay somewhere, and then the car goes to park before it comes back to work to pick me up. The car takes me home, then it goes back out to pick up my wife. You’ve generated all these empty-vehicle trips and empty-[vehicle miles travelled], more than if we had two individual vehicles.”

Of course, most of these problems are moot if the transportation system is built in the right way. A shared autonomous-vehicle fleet, like a ZipCar that comes to your door, could maximize the use of individual cars. A transit agency could work with a fleet of cars to direct riders directly to its stations, rather than letting them go elsewhere.

Couple that with growth in clean-fuel and electric vehicles—which is already being driven by federal fuel-economy standards—and concerns about pollution shrink.

“If you had asked these questions five years ago, you’d say this is some Silicon Valley pipe dream,” said NRDC’s Hwang. “But this tech is here and now and it’s moving so fast, that we need to work with industry and the government to steer it in the right direction.”

Earlier this year, House Republicans introduced a bill on automobile safety that included language that would have allowed automakers to use connected-vehicle technology to help meet fuel-economy standards (the regulations currently provide credits for technology such as electric vehicles to help automakers meet the standards, even if consumers keep buying gas-guzzling cars). Having that language would “slow the development of more efficient and less polluting technology,” environmental groups said in a letter asking that provision be struck.

“We’re not opposed to that technology, but … this is a completely different animal than getting a fuel-economy credit,” Hwang said. “We shouldn’t be giving credits when we don’t know how it’s going to play out.”

And that’s the problem forecasting the future of autonomous vehicles. Some features—such as lane assist and self-parking—are already in cars, but nobody knows how quickly that fully autonomous cars will penetrate the fleet, or how they’ll be used when they do. The Obama administration is only just starting its regulatory work on the cars, with new guidance expected in the next six months. Transportation agencies can only start to plan for the vehicles, but will have to adapt on the fly.

MacKenzie said that’s why it’s important to always be thinking about the doomsday scenarios and make sure that the technology doesn’t inadvertently undo the strides the auto sector has made on efficiency.

“We certainly can design around these possibilities, but it’s not certain that we will,” MacKenzie said. “There’s a lot of things that could happen, but I can’t say they will happen. It’s hard to know at this stage what is good policy, so I don’t know if there even is a good answer yet.”

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.