A new science of prediction

A new approach to predicting geopolitical and business events combines scenario planning for multiple alternative futures with forecasting methods that put hard probabilities on possible events to come.

Why it matters: Every policy is a prediction, as futurists like to say. Researchers hope a new science of prediction can improve the chances that policymakers and business leaders won't be caught off guard by rare black swan events, while allowing them to prioritize their preparations.

• "People are notoriously bad at anticipating the future, and countries aren’t much better," Foreign Affairs editor Gideon Rose writes in a new multipart package on the art and science of prediction.
• That much is plainly demonstrated by the failures of the U.S. and other countries to anticipate and respond to COVID-19, even though the last 20 years have featured multiple outbreaks of new pathogens, as well as countless reports and war games warning of just such a pandemic.

Context: Two main approaches have emerged to help us push through our psychological biases and develop actionable predictions about the future.

• Scenario planning, or strategic foresight, involves the creation of multiple possible imagined futures. When we face a situation that has no clear past analogs — like the dawn of nuclear weapons, which utterly changed the nature of warfare — scenarios can give us what futurist Herman Kahn called "ersatz experience" to help us plan for an imagined future.
• Forecasting involves seeing the future as a collection of discrete events, then using experience and reasoning to assign specific probabilities to whether those events will or will not occur.

The catch: Each of these techniques have their downsides and those drawbacks mirror each other.

• "The problem with forecasting is thus the exact opposite of the problem with scenarios," political scientist Philip Tetlock and prediction expert J. Peter Scoblic write in Foreign Affairs. "If the latter often provide too panoramic a view of the future to be useful, the former provides too narrow a glimpse."

How it works: The solution is an approach that synthesizes the best of both techniques, by crafting smart scenarios and using forecasting to create what Scoblic calls "meaningful signposts."

• For example, if we're trying to predict how COVID-19 will change the world, we can generate different scenarios: one where the virus is more rapidly brought to heel and life returns to a pre-pandemic normal, one where it endures for months or even years and unspools globalization, and so on.
• Each of these scenarios might be plausible, but there's no way for a president or a CEO to prepare for all of them. Forecasters can help by developing clusters of questions that "give early, forecastable indications of which envisioned future is likely to emerge," as Tetlock and Scoblic write.
• In the case of COVID, forecasters could assign a reasonable probability of whether an effective vaccine could be distributed by next spring. If forecasters are skeptical that will occur, we would be wise to plan for the long haul.

Yes, but: Humans are not good at identifying events in the moment that will end up driving the future, which is what we need for prediction.

• For example, last week the Nobel Prize in chemistry was awarded to the co-discovers of CRISPR, the gene-editing technique that everyone now understands will have a major impact on the future of humanity. Yet when the original paper describing CRISPR was published in 2012, journalists failed to recognize its significance, and no one wrote about it.

What's next: Algorithms can crunch vast amounts of data about the past and the present to identify trends, and Scoblic sees "human-machine hybrid forecasting as the next great question in anticipating the future."

• Notably, though, AI predictive models struggled during the early stages of the pandemic. The machines had been trained on normal human behavior, but COVID-19 put a temporary end to normalcy.

The bottom line: We are living through a moment of historic uncertainty, meaning it has rarely been as difficult to predict the future as it is now — and rarely as important to try to do so.