Deciding Under Deep Uncertainty: Choosing Well When You Can't Know the Odds

You already own the clean machinery of good decisions: weigh a bet by its expected value — probability times payoff — respect the fat-tailed world where one rare extreme runs the whole show, keep a margin of safety so a bad estimate doesn’t kill you, and shape your bets so the downside is capped and the upside stays open (asymmetry and optionality). This course is the honest sequel to all of it. It asks the question those tools quietly dodge: what do you do when you don’t know the probabilities at all — and can’t?

Expected value needs numbers. Multiply the odds by the payoffs, sum them up, pick the biggest. But that machine only runs if you can feed it the odds — and for the decisions that matter most (which technology wins the decade, whether this pandemic spreads, how a new market behaves, whether your one big plan holds up), the probabilities themselves are unknown, contested, or genuinely unknowable. Economist Frank Knight drew the line a century ago: risk is a gamble whose odds you know (a fair die, a roulette wheel), while uncertainty — what we’ll call deep uncertainty — is a gamble whose odds you don’t. Treat the second like the first and you get false precision: a beautiful spreadsheet, a confident single forecast, and a decision that shatters the moment reality lands in a cell you never modelled.

The expert move is a shift in the goal itself. When you can’t reliably compute which action is best, you stop hunting for the optimum and start demanding robustness: pick the action that does acceptably across many possible worlds rather than the one that looks best in your single guessed world. That one reframing unlocks the whole toolkit — satisficing instead of maximising, minimax-regret (minimise how much you’ll kick yourself in the worst case), and above all the ruin-avoidance / precautionary principle: never risk what you cannot afford to lose, because the game is played through time, and one ruin ends it — there is no average to come home to when you’re out of the game. You’ll buy back your ignorance with reversibility and option value (keep choices undoable, pay for margin of safety, run small experiments), with redundancy and slack (the antifragile spare capacity that turns a shock into a scratch), and with scenario thinking and pre-mortems that replace one brittle forecast with a spread of futures you’ve already rehearsed.

You’ll drive an interactive many-worlds explorer: pick a strategy — optimise, hedge, barbell, or robust — and drag the future from calm to deeply uncertain, watching the whole spread of possible worlds at once. The lesson lands hard and fast: in a calm world the optimiser wins on average, but crank the turbulence and it is the very strategy that gets wiped out, while the robust choice trades a little average return for surviving every world — and the barbell actually gains from the disorder. Along the way you’ll learn the traps that catch experts: false precision dressed as rigour, treating deep uncertainty as if it were mere risk, over-hedging into paralysis (paying so much for safety you never win), and mistaking robustness for plain pessimism. By the end you’ll hold the model that sits underneath every wise decision made in the fog — the meta-skill of choosing well precisely when you can’t calculate your way to the answer.