When to Trust Hunches — Summary

You are at the grocery store. You have cooked chicken three times this week. You know it will turn out well. You reach for fish instead.

An AI system might explain this as an explore-exploit algorithm: the information value of trying something new outweighs low confidence at low stakes. But that is the rationality reconstructed after the fact. What it felt like was simpler. It felt like choosing.

This matters because context determines when confidence is enough to act — and the thresholds are not fixed. Forty percent confidence justifies trying a new restaurant, because the stakes are low and the failure is reversible. Forty percent confidence does not justify surgery, unless the alternative is worse. The same number means different things depending on stakes, reversibility, time pressure, and what the cost of waiting looks like.

Hunches earn more trust when they come from domain expertise, when the track record supports them, when the stakes are low, or when time pressure makes deliberation impossible. The emergency physician who acts on incomplete information is not being irrational. Waiting for certainty in a deteriorating situation is the irrational choice.

Epistemic humility sometimes means acting despite uncertainty. Perfect confidence is not available. Waiting for it is paralysis. The question is not how confident you need to be before acting, but given your uncertainty, what is the wisest action.