Skip to content
Mental Models

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

Final Exam: Deciding Under Deep Uncertainty

A graded, one-way final exam on deciding under deep uncertainty — risk vs. Knightian uncertainty, why expected value lies (fat tails, false precision, the ergodicity trap and ruin), robustness over optimality, the buy-back toolkit, and the four traps. Pass mark 70%.

22 min Updated Jul 4, 2026

This is the final exam for Deciding Under Deep Uncertainty. It pulls the whole course together: Knight’s split between risk you can compute and uncertainty you can’t; the three ways expected value betrays you under deep uncertainty (the point estimate hides fat tails, the odds were invented, and the ergodicity trap turns a positive average into near-certain ruin for the one person living the sequence); the switch from optimality to robustness (satisficing, minimax-regret, and the ruin-avoidance gate); the buy-back toolkit (reversibility and option value, margin of safety, small experiments, redundancy, scenarios, the pre-mortem, the antifragile barbell); and the four traps — false precision, treating uncertainty as risk, over-hedging into paralysis, and mistaking robustness for pessimism. Several questions look easy until you spot a hidden trap: an invented probability dressed as a fact, a high average that hides a path through zero, or a “robust” plan that’s quietly just pessimism. Diagnose the situation before you answer each one.

Warning:

How this exam works

Read carefully — this exam is final. Each question appears one at a time. Once you submit an answer it is locked for good: there’s no going back, no retry, and no restart. Your score is hidden until the end, where you’ll see a pass/fail verdict. The pass mark is 70%. A few questions ask you to select all correct answers.

Question 1 of 24

In Frank Knight's distinction, what separates 'risk' from 'uncertainty'?

Select an answer to continue.

Course Recap

Big picture

Deciding under deep uncertainty, in one picture

  • Deciding Under Deep Uncertainty
    • Risk vs. uncertainty (diagnose)
      • Knight's split: risk = knowable odds (a die, a mortality table), optimise; deep uncertainty = unknown/unknowable odds (which tech wins the decade), switch modes. Optimising invented odds is false precision.
    • When expected value lies
      • The point estimate hides fat tails; the odds were invented; and the ergodicity trap turns a positive ensemble-average into near-certain ruin for the one person living the sequence (the 50/−40 coin: +5% mean, 1.5 × 0.6 = 0.9 downward compounding).
    • Robustness over optimality
      • Acceptable across many worlds beats best in one guess. Satisfice, use minimax-regret, and gate out ruin first — survival is the precondition, not a factor, because there is no compounding after zero.
    • Buy back recoverability
      • Reversibility and option value, margin of safety, small reversible experiments, redundancy and slack, scenario thinking, the pre-mortem, and the antifragile barbell so a wild world pays you.
    • The four traps + humility
      • False precision; treating deep uncertainty as risk; over-hedging into paralysis; mistaking robustness for pessimism. Underneath: calibration humility — you know less than your confidence suggests.
Success:

You've finished Deciding Under Deep Uncertainty

The expert model for choosing in the fog, in one line: diagnose whether you face risk or deep uncertainty; gate out ruin before you optimise anything; seek robustness — a choice that does acceptably across many worlds — over the fragile optimum for your single guess; buy back recoverability with reversibility, margin, experiments, slack, scenarios and the pre-mortem; and dodge the four traps while holding calibration humility. Expected value assumed you knew the odds. Now you know what to do when you don’t — stop optimising, and start surviving.

Mark lesson as complete