How does "Think in probabilities, not certainties" work?

Binary thinking — will/won’t, right/wrong — conceals the uncertainty that is actually present in almost every forecast. Expressing beliefs as probabilities forces intellectual honesty about confidence and creates a framework for updating: if you say 70% and the event happens 40% of the time, your calibration is off and you can improve it. Binary predictions are unfalsifiable until something fails completely.

How does "Run a premortem before committing" work?

Prospective hindsight — imagining a future event as if it has already occurred — has been shown to generate more and higher-quality reasons for why it might fail than forward-looking risk identification. The mechanism is that imagining the failure as real removes the wishful-thinking filter that blocks identification of uncomfortable risks when the decision has not yet been made.

How does "Check the outside view before trusting the inside view" work?

The inside view — the story you tell about your specific situation — systematically overweights vivid details and underweights base rates. The outside view — how comparable situations have actually turned out — is the corrective. Combining both is more accurate than either alone, but the inside view dominates by default, so the outside view must be actively sought.

How does "Update beliefs explicitly when new information arrives" work?

Motivated reasoning causes people to accept confirming evidence uncritically and scrutinize disconfirming evidence heavily. Explicitly updating a stated probability when new information arrives counteracts this: it forces engagement with whether the new information actually moves the probability and by how much, rather than allowing the old position to persist through selective attention.

How does "Find a truth-seeking group that argues with you" work?

Sycophantic feedback — people agreeing with you to preserve the relationship — is among the hardest biases to overcome because it feels good and arrives reliably from people who like you. A deliberately adversarial or truth-seeking peer group creates the kind of friction that surfaces the flaws in your reasoning before they cause a costly decision. The benefit is calibration from external challenges rather than from outcomes.

Annie Duke

What is thinking in bets and how does it improve decision making?

Probabilistic thinking, separating luck from skill, and building a better decision process

Short answer

Annie Duke's thinking in bets framework treats decisions as bets with uncertain outcomes, separating the quality of the decision from the quality of the outcome. Most real decisions involve incomplete information, so good decision-making is about process and probability, not about being right every time. The core skill is evaluating decisions on the information available at the time, not on how they turned out.

Annie Duke, a professional poker player turned behavioral scientist, argues that most of life's important decisions share poker's defining feature: you act on incomplete information and only find out the result later. The result is partly luck. Treating decisions as bets forces you to think probabilistically — what are the odds? — and to evaluate your choices on the quality of the reasoning at the time, not on whether you got lucky. Here are the core practices, with honest evidence on each.

The practices (6)

Why it works

Resulting — judging a decision by its outcome rather than the quality of the reasoning — is the central error the framework addresses. Outcomes are partly luck; the decision is the only thing you controlled. Systematically grading yourself on outcomes rather than process trains the wrong skill and distorts how you update your decision-making going forward.

How to do it

1After any significant decision plays out, write a brief evaluation: "was the reasoning sound given what I knew at the time?"
2Evaluate that separately from: "did it work out?"
3Look for bad processes that worked (lucky) and good processes that failed (unlucky) and treat them accordingly.
4Resist congratulating yourself for a good outcome that came from poor reasoning.

Evidence

Mechanistic

Resulting is well documented in hindsight bias research — people reconstruct decisions as obviously right or wrong based on outcome, obscuring the uncertainty that existed at decision time. The practice of process-based evaluation is recommended in forecasting and decision science but not itself studied in an RCT.

Honest caveat: Separating process from outcome is conceptually sound; doing it reliably in practice requires explicit record-keeping since hindsight bias operates automatically.

— Fischhoff (1975), hindsight bias and outcome knowledge, Journal of Experimental Psychology

Common mistake: Believing you can distinguish luck from skill retroactively without pre-commitment records — hindsight makes good luck look like skill and bad luck look like bad process.

Go deeper on this practice →

Practice this with IX Coach

Reading about a practice changes nothing on its own. IX Coach turns these into a guided, adaptive routine — discerning where you are in real time and walking the practice with you, session after session.

Start with IX Coach →