14 claims in “decision science”
In a betting frame, the winning strategy is to have the most accurate beliefs about the world, which means improving post-outcome reviews and ex ante probability estimates rather than being right about any single result.
A decision is informed by two boxes, stuff you know, which is beliefs with varying confidence, and stuff you don't know, which includes unknowable outcomes and information that is too costly to obtain.
Asking 'want to bet?' forces a person to consider the range of possible outcomes and their probabilities, which brings the influence of luck and their uncertainty to the surface.
When seeking input, present the options under consideration without revealing which one you favor, because revealing a preference causes others to either disagree or be persuaded without noticing.
Chess has full information and little luck, so the result reflects decision quality and resulting is reasonable there, while poker has hidden information and luck, so a single result can mislead and poker is a better model for decision making.
Calibrating beliefs can strengthen, weaken, or fully reverse them, and full reversals are the most valuable updates because a completely wrong belief distorts every decision that depends on it.
Understanding why an opponent makes a seemingly poor play, rather than dismissing them as an idiot, yields two benefits: incorporating reasonable strategies and recognizing the opponent's pattern in the future.
Some decision-relevant information is too costly in time or money relative to the improvement it would bring, and some people do not know what they do not know or what information would cost, which makes the decision analysis more complex.
Defining the game as accuracy means discovering one is wrong counts as a win, because it improves future decisions, which lets people reframe a short-term identity loss as a long-term gain.
Patrick argues that people focus too much on outcomes when judging decision quality and that they should instead focus on the process, which Duke's framework is designed to support.
After a decision is made, luck intervenes, so a decision creates a set of possible outcomes with probabilities and only one occurs, which means the outcome alone does not reveal the decision's quality.
In poker, uncertainty includes not only hidden cards but the unknown actions of opponents, so the skill is estimating ranges of opponent hands and how they will act, not just computing card odds.
To get high-fidelity feedback on a decision, describe the decision before the outcome is known, such as a trial lawyer reviewing strategy before the jury verdict, and never reveal the outcome to reviewers.
When the quality of a decision is transparent, such as a safe drive through a green light that is hit by a driver who ran a red light, people judge it as good despite a bad outcome, so making decision quality visible reduces outcome bias.