You are cruising down the highway at 65 miles per hour, reading a book in your self-driving car. Your life is in the hands of a machine—an eminently benevolent one. Meanwhile, in the lane next to you, an 18-wheeler using decidedly last-century technology—relying on a fallible human driver—appears to be swerving your way.

Your car’s computer is on the case. Equipped with orders of magnitude more computing power than the Apollo moon lander, it determines with all the confidence it can muster that there’s a greater-than-50-percent chance—it’s “more likely than not”—that the truck is about to hit you.

You may want to look up from your book. More importantly, you want to know with certainty that your onboard computer will hit the brakes, even if there’s a 49-percent chance that doing so will be a false alarm.

If, instead of “more likely than not,” the danger were “likely,” “very likely,” or even “extremely likely,” the answer would be clearer still. Even if there’s a 95-percent probability of a crash, there’s still a 1-in-20 chance that nothing will happen—but no one would gamble their life on those odds. Your car’s computer hopefully will have engaged the anti-lock braking systems already.

A perfect self-driving car doesn’t exist yet, nor has the world solved global warming. But it’s surprising that, by the standards that we’d expect in a car to keep its occupants safe, the governments of the world haven’t stepped on the brakes to avoid planetary-scale global warming disaster—a 100-year-storm hitting New York every other year, frequent and massive droughts, inundated coastal cities. In 1995, the Intergovernmental Panel on Climate Change declared that it was “more likely than not” the case that global warming was caused by human activity. By 2001, it had progressed to “likely.” By 2007, it was “very likely.” By 2013, it was “extremely likely.” There’s only one step left in official IPCC lingo: “virtually certain.”

But we can’t possibly want to wait for that. The 1995 declaration of “more likely than not” should have been plenty to convince us that keeping up the status quo—for two more full decades, by now—was the wrong call. Of course, the world’s systems of governance don’t operate as slickly as an electric engine or an automatic driving algorithm dreamed up by the world’s best engineers. They’re a veritable mess; the technical political-science term used in the context of the climate, "regime complex," doesn’t sound much better.

The problem at hand is similarly messy: There’s no single technology to rescue the world from hitting the global-warming wall. It takes channeling the actions of seven billion individuals who can’t be counted on to do the right thing voluntarily. Changing policy, in turn, takes votes: votes against a system that is pretending to be doing just fine speeding along at 65 miles an hour.

Opinions differ on whether the 1997 Kyoto Protocol that followed on the heels of the “more likely than not” determination has had a significant impact on global emissions. No one would say that it was close to enough. There are hopeful signs in some places, but if anything, the world has since been speeding up on its collision course rather than slowing down.

So how certain does science have to be for the world to act? If “more likely than not” was not enough to act in a way that is commensurate with the magnitude of the challenge, will “extremely likely” turn out to have been sufficient?

The language economists have developed to assess future unknowns can help give solid answers to these questions. Talk of "uncertainty" leaves itself open to uninformed calls to "wait and see." Even though "uncertainty" is technically worse than "risk," the latter is hands-down the more appropriate term, and economists know how to think about risk.

* * *

There are three points to be deduced from the overwhelmingly likely conclusion that humanity is causing climate change. Only one of them is positive.

The first piece of bad news is that we humans are, in fact, increasing global temperatures and sea levels alike. It would have been cause for celebration if, say, the 2013 IPCC report decided that science had gotten it wrong all along. But modern atmospheric science instead once again confirmed the basic ideas of high school chemistry and physics, going back to the 1800s: More carbon dioxide in the atmosphere traps more heat.

The good news, in some twisted philosophical sense, is the confirmation of bad news. Climate science has progressed over the past couple decades to the point where it is now possible to make the definitive statement that it is extremely likely that global warming is caused by human activity.

But there’s an additional piece of bad news: the false sense of security conveyed by repeating how certain science is. We know enough to act. Ignoring that reality, by now, would amount to willful blindness. But what’s not known has the potential to make things significantly worse.

We know, for example, that last time concentrations of carbon dioxide were as high as they are today, sea levels were up to 66 feet higher than they are now. Camels lived in Canada. That was a bit over 3 million years ago, during the Pliocene epoch, and today’s world obviously looks much different.

The best available climate models come close in their temperature projections to what the world experienced during the Pliocene, but they aren’t predicting sea levels of up to 66 feet higher. Nor do they predict camels wandering around Canada. Not now, nor hundreds of years from now. Why? Donald Rumsfeld would chalk it up to “unknown unknowns.” Nassim Nicholas Taleb’s term is “Black Swans.”

Economists have a way of thinking about projections that can help clarify how this sort of issue should be treated. We typically like to point out how Frank Knight, in his dissertation in the 1920s, at the University of Chicago, gave us the ironclad definitions of "risk" and "uncertainty." The former describes what it’s like to play a game of cards and not knowing which card will be next. The latter describes when you don’t know the distribution from which your card will be drawn. Harvard’s Richard Zeckhauser has added a third category: "ignorance," when you don’t even know what game you are playing.

Talk of “more likely than not” implies that we know the distribution, but more often than not, we don’t.

Meanwhile, most climate models are unduly skewed toward the known, sometimes making them much too conservative. Until recently, most climate models predicted rising sea levels only based on thermal expansion of the oceans (and the melting of mountain glaciers), but they did not include the effects of melting ice sheets. Warmer waters take up more space, leading to higher sea levels. That mechanism alone has indeed contributed to over a third of sea-level rise in the past two decades. It’s also clear that melting glaciers in Greenland and Antarctica raise sea levels, but by how much is highly uncertain. Call it a "known unknown." Until recently, scientific understanding of melting polar ice caps had been so poor that most models simply left it out. The IPCC did not include it in its projections until 2013.

Even though climate models do get a lot of things right, there are still lots of uncertainties built in. They’ll always be there, at least until well after it is too late to do anything about the problem in the first place. Uncertainties exist around the amounts of global warming pollutants we emit, the link between emissions and atmospheric concentrations, the link between concentrations and temperatures, the link between temperatures and physical climate damages, and the link between physical damages and their consequences. Also, at least as important, there’s uncertainty about how society will respond: what coping measures will be undertaken, and how effective they will prove to be.

Nailing down one of those steps—from concentrations to eventual temperature increases—has proven particularly elusive. The past three decades of advances in climate science have gotten us no closer to pinpointing the true answer. If the carbon dioxide concentrations in the atmosphere double—something that will surely happen, unless ambitious climate policies are enacted now—global average temperatures are likely to go up by between 2.7 degrees Fahrenheit and 8 degrees Fahrenheit. Scientists’ confidence in that estimate has increased, but what’s now called the “likely” range hasn’t changed since the late 1970s.

None of that means we don’t know enough to act—decisively, now. Yet it also means that climate is best thought of as a risk-management problem, one on a planetary scale. We need to look to the frighteningly high probabilities of even more extreme events and zero in on those for our benefit-cost calculus of when and how to hit the brakes.

Donald Rumsfeld, Nassim Nicholas Taleb, and anyone understanding what it means to manage risk would approve.


This post has been adapted from Gernot Wagner and Martin L. Weitzman's forthcoming book, Climate Shock: The Economic Consequences of a Hotter Planet.