How Not to Predict the Future
When you spend a lot of time thinking about emerging technology, you begin to notice certain things that are naturally just assumed to be true about the future, as if everyone just knows them. These ideas are so widespread and infrequently challenged that they come to seem inevitable.
These are also things that a lot of people really want to be true. What a coincidence.
So we take for granted that electric cars are the future, and the internal combustion engine is on the way out. That we're going to power everything with renewable energy, especially solar, and oil is a dinosaur (so to speak). And when self-driving cars are perfected, which will be any day now, they are going to put millions of blue-collar workers out of a job. But at least they will free up huge amounts of space inside cities, where we'll be able to replace ugly streets and parking lots with parks and playgrounds and affordable housing.
I hate to be the one to break it to you that a lot of these predictions are based on dubious assumptions and a dubious methodology. They survive and become widely accepted, not because they have been rigorously proven, but because they are "too good to check."
The field of meteorology has given us a very useful term for this: "wishcasting." It started with the observation that weathermen disproportionately predict sunny weather on the 4th of July and snow on Christmas Day. Their forecasts are influenced, not just by the evidence, but by what they (or their audience) want to hear. The term has since made the jump into politics, where it is used to describe the tendency of pundits to interpret the latest polls to show that their guy is winning and the other guy is fading, irrespective of the actual numbers.
It's no surprise that futurists tend to do the same thing, and often for the same motives: wish-fulfillment for their political preferences. Robots will be doing all of our work for us, they tell us, so we can finally afford to sit back and enjoy a generous "basic income" provided by the government. We will have everything powered with solar energy and batteries and electric cars, so we can finally shut down the oil companies without having to sacrifice our First World lifestyle. These are the fondest wishes of those with a certain cultural or political predisposition, so it is comforting to hear projections that technology will inevitably deliver these things.
(There is also a perverse mirror image of this, which is that there is always a market for dystopian predictions which tell us that technology is going to destroy us, as we always knew it would in the end.)
But a lot of these predictions depend on a very basic methodological error. They assume that new technology will change the way people live and work and think and make money—but only in one way. New technology will revolutionize life in one particular respect, while everything else will stay the same.
This, by the way, is also the leading cause of Zeerust, the datedness of old projections of the future. In the original "Star Trek" series, for example, we were going to figure out how to travel many time faster than the speed of light, but we would still be controlling all of our machines with kludgy Bakelite push-buttons from 1966. Or consider the old AT&T ads from the early 1990s, which speculated that one day you would use a tablet computer to send a fax from the beach. A lot of us own tablet computers, but when was the last time you sent a fax?
Similarly, we are shown the latest advances in the cost and efficiency of electric vehicles, and we are asked to project from this how quickly electric cars will surpass the internal combustion engine—but we assume (wrongly) that there are no advances being made in the internal combustion engine. Or we are told about the latest advances that will improve the economics of solar power, while fracking is quietly making its own advances that will make fossil fuels cheaper.
And let's take a closer look at that claim about self-driving cars and parking lots.
This is based on a straightforward calculation. When people own cars, they use them only about 4% of the time on average—to go back and forth from work and maybe to run a few errands. The rest of the time we're in our houses and offices, and the cars are sitting idle. So in theory there are 25 times as many cars as necessary. Well, it's not that simple. If everybody's 4% reliably came at different times of the day, a dozen people could share a single car between them with no problem. The problem is that most of them want to use their cars at pretty much the same times of day, before and after work. The reason we bother to own a car is partly to have the assurance of knowing it will be available on demand and that we won't have to wait (or pay "surge pricing") because we're competing with other users.
But let's say we account for this. All of these analyses assume that people will use self-driving cars for pretty much the same amount of driving as they engage in now. So these analyses assume that we can use driverless technology and ride-sharing software to divide fewer cars to cover the same overall distances.
That is, they engage in the kind of static analysis I was just talking about. They assume we can project the consequences of self-driving technology by assuming that people change their behavior in one way in response to new technology, but not in other ways.
But what if, when people no longer have to drive themselves, when you can spend your commute watching television or reading a book or getting an extra half-hour of work or an extra half-hour of sleep—that is, when time in the car no longer counts as time taken away from the rest of your life—what if people then want to do more of it.
The average person today has a commute to work of about 25 minutes. We can assume that this represents the average amount of time people are willing to give up out of their lives for daily travel. But what if your commute time becomes recreation time, or game time, or family time, or work time? Would people be more willing to, say, live much farther out of town—where houses are cheaper and yards are more spacious—and enjoy a nice, leisurely hour-long commute? And when it comes to travel for pleasure, I suspect you would be more willing to drive two hours on a daytrip if you didn't actually have to do the driving.
And if people are doing all of these things in their cars, would they be happy to do them in a generic model sent over by a ride-sharing service—or would they be even more eager to own their own car, customized to their specific desires and needs? Would they want to build themselves self-driving offices or self-driving living rooms, or even self-driving bedrooms?
Consider what this means for the projected gains in fuel efficiency, or for expected decreases in fossil fuel consumption, from self-driving cars and ride-sharing. If people want their self-driving cars to perform all of those other functions—office, living room, bar, bedroom—will they actually want them to be bigger? Look at the ads for automobiles. Most of them advertise how fun it is to drive the car. But what if automakers are no longer competing on the basis of the driving experience? They might compete instead on luxury and spaciousness and creature comforts. That Google/GM/La-Z-Boy collaboration can't be far off.
And if the vehicle ends up being bulkier and more difficult to maneuver, what do you care? You won't be driving it, you'll be lounging in the back with your feet up. If we come to view our car as a living space, and not as something we drive, we might find ourselves looking, not for a self-driving Prius, but for a self-driving Winnebago.
The same thing goes for speculation that self-driving cars and trucks will eliminate millions of jobs for drivers and truckers. Maybe so, but that also assumes that our demand for goods and services remains unchanged. This is the basic error of all previous predictions about machines eliminating human employment. It assumes that we will change one aspect of our way of doing things—say, the machines we use to weave cloth—but that the amount of cloth we want will remain exactly the same. So when the new machines hit the same old quotas, that's the end of human employment. What has actually happened was that the new machines gave us the same things at cheaper prices, and suddenly we found that we want more things and new things, and people shifted their employment over to producing those new things.
If we want some lessons for the adoption of the driverless car, we should look to the adoption of the horseless carriage.
I recently had reason to look up the distance the average person drives in a year. It's about 15,000 miles. What immediately struck me was: before the automobile, how many people traveled 15,000 miles in a year? That distance would have seemed extraordinary. Only a handful of diplomats and adventurers (and maybe sailors and railway workers) would have been able to claim it, and even fewer would have been happy about it, since that amount of travel was generally time-consuming and arduous.
In fact, before the automobile, there were probably many people who never traveled 15,000 miles in their entire lives. Where I live, in a rural area of Virginia, I've noticed that there are a lot of old place names that belong to locations that aren't really places any more. Along any given rural highway, you will notice that every five miles there is a little sign with the name of an old town, around which there is perhaps a cluster of old houses or a noticeable widening of the road, but not much more. Yet the old place names tell you why they were there: Boyd Tavern, Zion Crossroads, Kent's Store, Brandy Station. You can also understand why they were dotted around the countryside every five miles. This reflects the average distance people were willing to travel by foot or by horse on a regular basis.
At that time, many people would have traveled less than ten miles a week in the ordinary course of their lives, at which rate it would have taken them three decades to travel the distance the average American covers by car in a year.
The point is that when you introduced the horseless carriage, a futurist of the era might have predicted that it would save people an enormous amount of travel time, since they could now cover in a few minutes a distance that used to take an hour. That's the prediction he might have made if he assumed people would use the new technology without altering their old behavior. But people actually chose, not just to cover the same distances faster, but to cover much greater distances.
Maybe that won't happen. Maybe self-driving cars won't lead people to want longer commutes or to own bigger cars. Maybe they will gravitate toward small, shared, electric vehicles. The actual result will undoubted be a mix of these things, but we don't yet know in what direction that mix will be tilted. And we won't know for decades.
Which calls for us to be more cautious and less confident in our speculations. The basic paradox of futurism is that the very thinking that makes projecting the future interesting is the thing that makes it perilous. When we try to project the impact of future technology, we're thinking about something that doesn't exist yet, so its exact implementation and uses are open-ended unknowns. That's exciting, because it gives us a lot of scope to explore new possibilities. But for precisely that reason, it also means that a lot of our speculations are going to be wrong.
That would be fine if it were just a matter of idle speculation. The problem is that cities are being asked to plan around the presumed effects of self-driving cars. The futurists' wishcasting about fewer roads and parking lots and more parks and playgrounds is just another repetition of the dream urban planners have had for cities pretty much since the invention of the automobile. But the usual vice of urban planners is that they try to impose their pre-existing priorities onto people who may or may not want to live in their ideal city. So we see that a lot of these articles about the self-driving future include plans to nudge people into using self-driving cars the way the planners want them to be used. But people don't like to be nudged, and they resist attempts to make them live the way someone else wants them to live.
After all, what's the point of a disruptive technology if it doesn't disrupt the plans of the planners, too? It's the same error: futurists want technology to disrupt only the select few things that they think need disrupting, while it leaves everything else exactly the way it was.
The risk is that, fifteen years from now, whole new fleets of self-driving cars will be trying to fill up the streets and parking lots that city planners have already earmarked for their other goals.
There is certainly a purpose in asking what would happen if a technological disruption goes all the way in a certain direction, as a kind of thought experiment. And sometimes the extreme possibilities come true, or close to it. How many people predicted, for example, that the phone call would be in danger of becoming obsolete?
But we should also maintain some skepticism about how likely we really are to get it right—particularly when our forecasts fall in just a little too well with our wishes.