If We Knew the Future, Would We Know What to Do?

The art of thinking ahead

Forecasting, intelligence, and the difficulty of acting in time

When I think about the relationship between intelligence and political decision-making, a small thought experiment often comes to mind. Imagine handing a government next year’s newspaper today. How many of the stories would be taken seriously? More importantly, how many would prompt a decision before it was too late? It is a paradox, of course, but it points to a difficulty we tend to underestimate: having information about the future, or at least reliable indications of what may happen, does not mean knowing how to use it.

I returned to this question while rereading A Better Crystal Ball: The Right Way to Think About the Future, an article published in Foreign Affairs in October 2020 by J. Peter Scoblic and Philip E. Tetlock. Their starting point is deceptively simple: every policy decision contains a forecast. When a government imposes sanctions, it assumes those measures will influence the target’s behaviour. When it invests in defence or cybersecurity, it is making a judgment about the risks it expects to face. Yet foreign policy decisions are made with incomplete information, and we often approach new problems using categories developed to understand old ones. History is indispensable to understanding the present. It becomes a trap when we treat it as a script for the future.

Scoblic and Tetlock propose bringing together two methods that often operate separately. The first is scenario planning: imagining several plausible futures to challenge settled assumptions and prepare for unexpected developments. The second is probabilistic forecasting: asking precise questions, estimating the likelihood of an event occurring by a given date, and revising that estimate as new evidence becomes available. Scenarios can turn into compelling stories without telling us how likely they are or when action is needed. Probabilities, meanwhile, can create a misleading impression of precision, particularly when applied to long-term strategic change.

The article’s most interesting proposal is to connect the two through clusters of testable questions, or question clusters. Suppose an intelligence service is assessing the risk of escalation in a particular region. Alongside its broader scenarios, it should identify observable developments in advance: the closure of specific diplomatic channels, the movement of identifiable military units into a defined area, or the introduction of particular economic restrictions. Each question needs a deadline, a reasoned probability estimate, and clear criteria for deciding whether the event has occurred. Analysts must also look for evidence that would undermine their preferred scenario, rather than collecting only signs that support it. This makes assumptions testable and gives institutions an opportunity to revise their judgments before a crisis unfolds.

This is where the argument becomes especially relevant to intelligence work. A good intelligence report should do more than describe what might happen. It should help its readers understand which developments matter, which signals deserve attention, and what might be lost by delaying a decision. In its opening paragraphs, I would want to see a few clear judgments: the main scenarios, the evidence that makes them more or less likely today, the remaining uncertainty, and the practical consequences of different courses of action. The rest of the report should make those judgments open to scrutiny by setting out sources, assumptions, and possible objections. Such a report needs regular updating, along with a willingness to acknowledge when a forecast was wrong.

None of this means turning the analyst into a political decision-maker. Their responsibilities must remain distinct: intelligence assesses facts, risks, and potential consequences; setting political objectives and accepting political responsibility belong to those who govern. That is precisely why analysis should clarify the available options, including the costs of inaction, without presenting any one choice as inevitable. A 70 percent probability, with no explanation of the evidence behind it or how it will be updated, may be less useful than a candid qualitative assessment of what remains unknown.

Scoblic and Tetlock themselves argue that the greatest obstacle is organisational. Institutions are consumed by immediate demands, whereas anticipating a crisis requires time, continuity, and a willingness to question established judgments. I find this particularly relevant when thinking about European and national administrations. We can produce sophisticated analyses, multiply indicators, and refine forecasting models. Yet if no one knows when to revise a priority, change a plan, or take an unpopular decision, even excellent analysis may remain trapped in a well-written document.

I keep coming back to that newspaper from next year. The real challenge is not getting hold of it. It is building institutions capable of recognising the signals already in front of them, early enough to act on them, even when doing so is politically uncomfortable.

References