Decision-Making Under Uncertainty

Overview

Every important decision you will ever make is made with incomplete information. Waiting for certainty is not patience — it is a decision to do nothing, made by default. This conversation is about the skill that separates capable adults from passengers: choosing well when you cannot know, and being able to defend the way you chose even when the result disappoints you. It drives at a single uncomfortable idea — that a good decision and a good outcome are different things, and confusing them will quietly ruin your judgment.

This is a discussion, not a lecture, and it works best anchored to a decision the student is actually facing right now. Abstract decision theory is forgettable. The question "should I take on this client even though I am not sure I can deliver?" is not.

Why does this matter so much at this stage specifically? Because the student is, for the first time, making decisions whose consequences land on them rather than on a parent or a teacher. A bad grade is recoverable and someone else absorbs most of the cost. A signed contract you cannot fulfill, a product launched into a market that does not want it, a partnership formed with the wrong person — these are decisions with teeth, and the person who learns to think about them well before the stakes get even larger has a compounding advantage for the rest of their life. The habits built here do not stay in the venture. They show up in every choice about money, relationships, health, and time that the next sixty years will demand. The aim of this conversation is to install a way of deciding that the student will still be using when the decisions are about millions of dollars and other people's livelihoods.

The Big Question

When you cannot know how things will turn out, how do you decide well — and how would you even know afterward whether you decided well?

There is no clean answer. The point of the conversation is to build a way of thinking, not to reach a verdict.

Context for the Facilitator

You are the mentor here, and the temptation will be enormous to tell the student what to decide. Resist it completely. Your job is to interrogate their reasoning, not to supply a conclusion. If they leave with your answer, the unit failed. If they leave with a sharper way of generating their own answers, it succeeded.

A few ideas anchor this discussion, and you should hold them well enough to push on the student's thinking:

Resulting. This is the central error, named by the poker player and decision researcher Annie Duke. Resulting is judging the quality of a decision by how it turned out. A drunk driver who makes it home safely made a terrible decision with a good outcome. A founder who researched carefully, raised capital, and still failed because the market shifted may have made a good decision with a bad outcome. Because we only ever see outcomes, we constantly mislabel our decisions — and then learn the wrong lessons. The fix is to evaluate the decision against what was knowable at the time, not against what happened.

Expected value. When outcomes are uncertain, the right comparison is not "what is the best case" or "what is the worst case" but "what is the average result if I made this same bet many times." A choice with a 70% chance of gaining 100 and a 30% chance of losing 50 has an expected value of (0.7 × 100) − (0.3 × 50) = +55. You will not get +55 on any single play — you will get +100 or −50. But over a lifetime of decisions, the person who consistently takes positive-expected-value bets wins, and the person who chases the rare big payoff usually does not.

Reversibility. Jeff Bezos drew a useful line between one-way doors (decisions that are hard or impossible to undo) and two-way doors (decisions you can walk back). Two-way doors should be made fast, with limited information, because the cost of being wrong is small — you reverse it. One-way doors deserve real deliberation. Most people get this exactly backwards: they agonize over reversible choices and rush the irreversible ones.

The premortem. Before committing, imagine it is a year later and the decision has failed completely. Then ask: what went wrong? This trick, from psychologist Gary Klein, defeats overconfidence by giving people permission to voice doubts they would otherwise suppress. It surfaces risks that optimism hides.

Optionality and the cost of waiting. Not deciding is itself a decision, and it has a price. Sometimes the price of waiting is small and the information you gain by waiting is large — in which case delay is the smart play, and you should hold the option open. But often the option to decide is itself decaying: the client will hire someone else, the window in the market closes, the co-founder commits elsewhere. A capable decider tracks two things at once — what they would learn by waiting, and what they would lose by waiting — and acts when the second outweighs the first. The student will be tempted to treat "I'll wait and see" as the safe, cost-free default. It is neither. Help them put a price on the waiting, in the same units (time, money, opportunity) as the action itself.

Base rates. Before reasoning about the specifics of a decision, ask what usually happens to people in this situation. How often do first-time ventures of this kind succeed? How often do clients who haven't signed actually pay? The specifics feel more relevant than the statistics — my situation is different — but the outside view, the base rate, is usually a better predictor than the inside story we tell ourselves. The student will resist this; everyone does. Their case always feels exceptional. Make them state the base rate out loud first, then argue for why they are or are not an exception. That ordering is the whole discipline.

You should also be ready to name the biases that distort decisions, because the student will exhibit them live during the conversation. Confirmation bias (seeking evidence that supports what you already want). Sunk-cost fallacy (continuing because of what you have already spent, which is gone regardless). Availability bias (overweighting whatever is vivid or recent). Overconfidence (systematically underestimating how wrong you might be). Catch the student doing one of these in real time and name it gently — that is worth more than any definition.

Opening

Open with a story, not a definition. Tell it like this, or tell one of your own:

"Two founders each start a company. One spends a month on careful research, validates the market, raises a modest amount of capital, and builds deliberately. The market shifts unexpectedly six months in and the company fails. The other founder has an impulse, maxes out a credit card, builds in a weekend, and gets lucky on timing — the company takes off. A year later, everyone calls the second founder a genius and the first one a cautionary tale. Who actually made the better decision?"

Let the student sit with the discomfort. Most people's instinct is to credit the winner. The whole conversation lives in resisting that instinct.

Discussion Guide

Phase 1: Surface Understanding

• In the founder story, separate the two things: who made the better decision, and who got the better outcome? Are they the same person?
• When you have made a decision that turned out badly, how did you judge yourself afterward — by the choice or by the result? Be specific. What was the decision?
• What is a decision you are facing right now where you do not have enough information to be sure? Describe it. We will use it for the rest of this conversation.

Phase 2: Dig Deeper

• For the decision you just named: what would you have to know to be certain? Now — realistically, can you ever know that before you have to decide? If not, waiting for certainty is just choosing not to act. Is that the choice you want?
• If you made this exact decision a hundred times, what fraction of the time do you think it works out, and how big is the win versus the loss? Walk through the expected value out loud. Does the math change your gut feeling?
• Is this a one-way door or a two-way door? If you are wrong, how hard is it to reverse? Does that change how much deliberation it deserves?
• Run a premortem with me. It is a year from now and this decision failed badly. Tell me the story of what went wrong. What does that story reveal that you were not letting yourself see?

Phase 3: Apply

• Where in your reasoning just now were you only looking for evidence that you are right? What evidence would actually change your mind — and have you gone looking for it?
• Is any part of your thinking driven by what you have already invested — time, money, ego — rather than by what is best from here forward? That money and time are gone whether you continue or not. Decide from where you stand now.
• How confident are you, as a percentage? Now: think of three specific things that could make you wrong. Has your number moved? Most people's overconfidence does not survive being asked to name the failure modes.

Phase 4: Synthesize

• Given everything we have said — expected value, reversibility, the premortem, the biases you caught yourself in — what is your decision, and what is the process reason you can defend it by, independent of how it turns out?
• Write the decision down before you act: what you are choosing, why, what you expect to happen, and how confident you are. This is your decision journal. Months from now you will reread it and learn whether your process was sound — separate from whether you got lucky.

A Worked Walkthrough

It helps to see the whole framework applied to one decision, so you know what a good run of this conversation actually produces. Suppose the student is a freelance designer and the decision is whether to take on a large, well-paying client whose project is slightly beyond their current skill — they think they can deliver, but they are not sure.

A weak version of this conversation ends with "go for it, you'll figure it out" or "don't risk your reputation." Both are verdicts, and both skip the thinking. Here is what the strong version produces, moving through the lenses in order.

Resulting check. First, detach the decision from the result. If they take it and it goes badly, that does not automatically mean it was a bad decision — and if they take it and pull it off, that does not automatically mean it was a good one. They are choosing under uncertainty either way. The question is whether the bet is good given what they know now.

Expected value. Make them do the math out loud. Say the project pays enough to matter and there is, honestly, a 65% chance they deliver well, a 25% chance they deliver something acceptable but stressful, and a 10% chance they fail and damage the relationship. What is each branch worth — in money, in reputation, in skill gained? The 65% branch is a large win on all three. The failure branch is the one to size carefully. Is the 10% downside a recoverable bruise or a reputation-ending event?

Reversibility. Is this a one-way door? Usually not entirely — they could scope the project in stages, take a smaller piece first, or build in a checkpoint where either party can walk away. A decision that looks like a single terrifying yes/no is often a two-way door in disguise once you look for the staging. Finding that staging is frequently the actual answer the conversation should produce.

Premortem. It is a year later and the project failed. Why? Almost always the honest answer is something specific and addressable — "I underestimated how long the unfamiliar part would take and missed the deadline." Good. Now they can de-risk that exact thing in advance: build a buffer, prototype the hard part first, set the client's expectations early.

The decision and the defensible reason. The output is not just "yes" or "no." It is something like: "Yes, in staged form, with the hard component prototyped in the first week as a kill-switch, because the expected value is strongly positive and the worst case is recoverable as long as I cap my exposure with the staging." That sentence is a process they can defend regardless of how it turns out. That is the deliverable of this conversation, every time: not the answer, but a defensible reason for the answer.

Facilitation Tips

• If the learner says "I don't know": Good — that is the honest starting point for every real decision. Do not let them off the hook, but reframe: "You do not need to know. You need to make the best bet with what you have. What is your best estimate, even if you are unsure?" Forcing a probability estimate out of "I don't know" is the entire skill.
• If the discussion gets heated: It usually gets heated when the student is defending a decision they are emotionally attached to. Name it without judgment: "I notice you are arguing hard for this. Is that because the reasoning is strong, or because you have already decided and want it to be right?" Then drop it and let them sit with the question.
• If they give a surface answer: Push for a number or a specific. "It'll probably work out" is not an answer. "What is 'probably' — 60%? 90%? And what is the size of the loss if you are in the other 40%?" Vagueness is where bad decisions hide.

Common Perspectives

Related Readings or Media

• Thinking in Bets by Annie Duke — the clearest treatment of resulting and decisions-as-bets. Read it after this conversation.
• Superforecasting by Philip Tetlock — evidence that ordinary people can become genuinely good at probabilistic forecasting through deliberate practice.
• The "regret minimization framework" — search for how Jeff Bezos described deciding to start Amazon. A short, practical lens for one-way-door decisions.
• Thinking, Fast and Slow by Daniel Kahneman — the foundational catalog of the biases named above. Dense; read it in sections, not at once.

Follow-Up

• Journal prompt: Start a permanent decision journal. For your next five significant decisions, before acting, write: the decision, your reasoning, your predicted outcome, and your confidence as a percentage. Revisit each one after the result is known and ask only one question — was my process sound, regardless of how it turned out?
• Action: Make the decision you brought to this conversation, and make it on the basis of process you can defend. Do not let it drift into a decision-by-default.
• Revisit in: Three months — reread the journal together and look for the pattern of resulting in your own self-judgment. That pattern is the thing to break.