How do you plan for the unpredictable? That's the question facing the more than 400 metropolitan planning organizations (MPOs) that have been tasked by Congress to write 20-year transportation plans for their regions. Self-driving cars will be on the market in the next 10 years, are likely to become a dominant form of travel in 20 years, and most people think they will have huge but often unknowable transformative effects on our cities and urban areas. Yet not a single regional transportation plan has tried to account for, and few have even mentioned the possibility of, self-driving cars.

Instead, many of those plans propose obsolete technologies such as streetcars, light rail, and subways. Those technologies made sense when they were invented a hundred or so years ago, but today they are just a waste of money. One reason why planners look to the past for solutions is that they can't accurately foresee the future. So they pretend that, by building ancient modes of transportation, they will have the same effects on cities that they had when they were first introduced.

If the future is unpredictable, self-driving cars make it doubly or quadruply so. Consider these unknowns:

• How long will it take before self-driving cars dominate the roads?
• Will people who own self-driving cars change their residential locations because they won't mind traveling twice as far to work?
• Will employers move so they can take advantage of self-driving trucks and increased employee mobility?
• Will car-sharing reduce the demand for parking?
• Will carpooling reduce the amount of vehicle miles traveled (VMT), or will the increased number of people who can "drive" self-driving cars increase VMT?
• Will people use their cars as "robotic assistants," going out with zero occupants to pick up groceries, drop off laundry, or do other tasks that don't require much supervision?
• Will self-driving cars reduce the need for more roads because they increase road capacities, or will the increase in driving offset this benefit?
• Will self-driving cars provide the mythical "first and last miles" needed by transit riders, or will they completely replace urban transit?

Planners from the MPOs in Seattle and Atlanta asked participants at the recent Autonomous Vehicle Symposium to help them incorporate self-driving cars in their regional transportation models. Yet the consensus was that no one has any idea about the answers to the questions I aksed above. The only prediction that people could come close to agreeing on was that self-driving cars will increase miles of driving as people take advantage of greater mobility more than they increase carpooling.

Self-driving cars are not a black swan amidst the flock of knowns about urban planning; they are a whole flock of black swans, any one of which could completely sink even the most accurate predictions about all the others.

Some of the planners believed they could make guesses about the effects of self-driving cars and use them to make "sensitivity runs" to estimate the possible magnitude of the effects of self-driving cars on cities. But even if they made such runs, they would have no idea which runs will come close to reality.

"There are no models in planning practices that can predict the emergence of new modes and forms of mobility," admitted one planner. "Our models haven't even got the Internet yet. They haven't got the cell phone. They're not going to have autonomous cars." Another agreed: "ITS [intelligent transportation systems] is 25 years old, but our models still don't account for it."

We are about to introduce a new technology that will completely transform our society in unpredictable ways, and many of those transformations will start changing travel behaviors and land-use patterns well before 20 years are up. The fact that the plans are revised every five years doesn't help because many of those plans include costly investments in projects that take decades to complete. Even if new information reveals that those investments are no longer appropriate, once begun the political pressure to complete the projects will likely be too great for future officials to resist.

This means it's not enough to simply rewrite transportation planning models. We need to rewrite the entire process of urban planning, following four principles:

1. Instead of writing 20-year plans that pretend to know what a city will need in the distant future, planners should only write short-term plans that solve today's problems without foreclosing options for the future.
2. Planning processes should be streamlined so that it no longer takes 10 or more years to plan, design, and build facilities that, a few decades ago, were built in a couple of years.
3. Urban areas should avoid infrastructure projects that take decades to complete and would make sense only if people completely changed their lifestyles.
4. New transportation facilities should be "generic" in the sense that they can be used by a wide variety of modes and easily adopted for whatever modes become dominant in the future.

If some of these suggestions sound familiar, that's because I've made them before, particularly in my 2007 book The Best-Laid Plans. The future is unpredictable even without self-driving cars, and I've had little faith in the ability of long-range plans to cope with those unpredictabilities. But now even the planners are willing to admit that they can't cope with the unpredictable effects of this new technology. I hope that at least some of them are willing to tell that to Congress, which created the requirement for 20-year plans, that it needs to change the rules.