Choosing Under Uncertainty / Decision-Making

Risk is not uncertainty

There is a distinction the language blurs and good decisions depend on. Risk is when you face the unknown with known odds — a roulette wheel, an actuarial table, anything where the probabilities are themselves known quantities. Uncertainty, in the strict sense Frank Knight gave it, is when the odds themselves are unknown, and sometimes unknowable: novel situations, complex systems, the long-term consequences of a choice no one has made before.

The casino runs on risk. Life runs mostly on uncertainty. And the tools that work beautifully for the first fail quietly and badly for the second.

Why expected value runs out

The standard machinery of rational choice says: multiply each outcome’s value by its probability, sum, pick the largest. Under risk, this is exactly right. Under deep uncertainty, it has a hidden dependency — it needs the probabilities, and under uncertainty you do not have them.

When the inputs are unknown, an expected-value calculation does not become impossible; it becomes false precision. You plug in a guessed probability, turn the crank, and out comes a confident number whose authority is entirely borrowed from arithmetic and not at all from knowledge. The decision feels rigorous. It is rigor applied to fiction. The most dangerous decisions are often the ones dressed in the most convincing quantification of things no one actually knows.

A precise answer built on an invented probability is not more reliable than an honest guess. It is less — because it has hidden the guess.

The shift from prediction to robustness

If you cannot know which future will arrive, the intelligent response is to stop trying to bet on one. The question changes from which outcome is most likely? to which choice holds up across the widest range of outcomes, including the ones I haven’t imagined?

This is the deepest move in deciding under uncertainty, and it inverts the amateur’s instinct. The amateur forecasts harder, seeking the one true prediction. The professional, knowing the forecast is fog, designs the decision to survive being wrong about it. You are no longer optimizing for the expected future. You are optimizing for resilience across possible futures — accepting a lower payoff in the world you’d bet on, in exchange for not being destroyed in the worlds you wouldn’t.

Robustness is not timidity

This is easy to mistake for mere caution, and it is not. Robustness is not refusing to act; it is acting in a way whose failure modes you can survive and whose success modes you stay exposed to. Sometimes that means hedging. Sometimes — as the next essay argues — it means the opposite: taking many small, survivable risks precisely because you cannot predict which will pay, and arranging things so the payoff comes anyway.

The common thread is humility about prediction converted into structure in the decision. You do not need to know the future if you have chosen something that does not depend on knowing it. That is the whole art: not better forecasting, but choices that are forecast-proof.

What actually matters

Once prediction is dethroned, attention moves to the features of a decision that determine whether it survives uncertainty — and these turn out to have almost nothing to do with the content of the choice and almost everything to do with its shape.

How are the payoffs distributed — is the downside bounded and the upside open, or the reverse? Can the decision be undone, or is it a door that locks behind you? What does it cost to be wrong, versus what it costs to be cautious? None of these are questions about predicting the future. They are questions about the structure of the bet. The next three essays take them one at a time, beginning with the most important: the shape of the payoff.