Superforecasting — The Art and Science of Prediction

Philip Tetlock’s first major book, Expert Political Judgment (2005), established one of the most uncomfortable findings in social science: expert forecasters — political scientists, economists, intelligence analysts — are, on average, no more accurate than dart-throwing chimpanzees when predicting complex world events. The experts who appeared most confident and were most sought after by media and government were often the least accurate. Superforecasting, written with science journalist Dan Gardner, describes what Tetlock discovered when he ran a follow-up study — a massive forecasting tournament called the Good Judgement Project — and found that a subset of ordinary people consistently outperformed professional intelligence analysts, even those with access to classified information.

These “superforecasters” were not especially educated or credentialed. What distinguished them was a set of thinking habits: they were actively open-minded rather than committed to a specific worldview; they sought out disconfirming evidence deliberately; they expressed their beliefs as specific probabilities rather than directional intuitions; they updated their forecasts readily when new information arrived; and they maintained a healthy scepticism about their own reasoning, what Tetlock calls “foxlike” thinking (drawing on Isaiah Berlin’s metaphor of the fox who knows many things vs. the hedgehog who knows one big thing deeply). The hedgehog experts — the ones with grand unified theories — were systematically worse forecasters than the foxes.

The practical framework Tetlock extracts from superforecaster behaviour is built around two concepts: the outside view and calibrated uncertainty. The outside view means starting any forecast by identifying the relevant reference class — how often does this type of event occur? — before adding case-specific details. This discipline resists the natural tendency to treat every situation as unique and to weight recent, vivid information too heavily. Calibrated uncertainty means expressing beliefs as percentages rather than categorical claims, and tracking those percentages against outcomes to measure your accuracy over time (via Brier scores). Forecasters who do this consistently improve; those who deal in vague confident assertions do not.

Superforecasting is more rigorous than most books in its genre and more empirically grounded than virtually any self-help adjacent title that makes claims about improving thinking. Tetlock’s research is genuinely consequential — it influenced how IARPA (the intelligence community’s research arm) and later several governments approach forecasting practice. The lessons transfer directly to business contexts where probabilistic forecasting about product launches, market conditions, or hiring decisions replaces gut-feel judgement. The most useful single habit the book teaches is deceptively simple: instead of saying “I think this will happen,” say “I think there’s a 65% chance this will happen,” and then check whether you were right 65% of the time across many such predictions.