Philip Tetlock

For twenty years Philip Tetlock ran the most rigorous study of expert prediction ever conducted, and the headline finding was devastating: the average credentialed expert, asked to make specific, time-bound, probabilistic forecasts in their own domain, performed no better than a dart-throwing chimpanzee. The punchline became famous. The second finding—that a small minority of forecasters consistently outperformed both the average expert and sophisticated statistical models—became the more important one. What distinguished these superforecasters was not intelligence or knowledge but cognitive style: the fox who knows many things, holds multiple frameworks simultaneously, treats its own confidence as a variable to calibrate rather than a virtue to defend. The hedgehog, certain and narrative-driven, consistently underperformed. In the AI age, Tetlock's lens becomes essential: the systems that now produce confident, fluent output regardless of accuracy have created an environment in which calibrated uncertainty is simultaneously more necessary and more threatened than at any previous moment in the history of human judgment. His own trajectory