Superforecasting

Editorial review

Built on a multiyear forecasting tournament run for IARPA, Tetlock and Gardner show that some ordinary people consistently outperform intelligence analysts at predicting world events. The habits they share are learnable and surprisingly mundane.

AI-generated summary

After his earlier work showed expert pundits performed barely better than chance, Philip Tetlock studied a large pool of volunteer forecasters to identify the cognitive habits of those who reliably outperformed: 'superforecasters.'

Key takeaways

• 1

  Probabilistic thinking is a learnable skill, not an inborn trait.

• 2

  Updating in small increments beats both stubbornness and flip-flopping.

• 3

  'Foxes' (many small models) tend to forecast better than 'hedgehogs' (one big model).

• 4

  Aggregated, calibrated forecasts can rival expert intelligence assessments.