What you'll learn
Key ideas from Noise
These ideas compress the book's argument without treating the author's view as settled fact. Use them as an orientation before reading the full work or listening in Wiseley.
Bias shifts judgments in a common direction, while noise scatters them.
The system, level, pattern, and occasion taxonomy applies across sentencing, business, medicine, government, and other judgment systems.
Averaging independent judgments reduces random noise, and delayed or self-critical second judgments can recover some of that benefit within one person.
Fixed procedures often improve prediction by applying the same information and weights without personal or occasion noise.
Outside views, base rates, regression, and calibrated humility keep individual cases connected to broader statistical reality.
Sequenced information, independent aggregation, shared scales, and structured subtasks form the core process safeguards of decision hygiene.
The final framework selects rules, standards, algorithms, guidelines, aggregation, or structured processes according to task, error, and institutional values.
Inside Noise
Read the first chapter in full here. The other 15 continue in the Wiseley app.
Chapter 1 of 16 · 4 min · Audio & text
Noise Hides in Plain Sight
Noise, by Daniel Kahneman, Olivier Sibony & Cass R. Sunstein.
Picture four teams shooting at a target. Team A's shots cluster around the bull's-eye: accurate and consistent. Team B's shots cluster together but miss it in the same direction: biased. Team C's shots spread widely around the target, with no obvious common direction: noisy. Team D is both scattered and off center. The image separates two forms of error. Bias pushes judgments systematically; noise makes them vary. Either can exist without the other, and a system can suffer from both.
Noise is easy to miss because disagreement does not announce which judgment is wrong. Yet scatter itself can be measured. We may not know the bull's-eye, the correct diagnosis, or the eventual outcome, but we can still see that qualified decision makers reach different conclusions. This matters wherever people make judgments on behalf of an institution. Doctors can disagree about a diagnosis, asylum judges can reach very different admission decisions, and child-protection workers can differ in whether they remove a child. If the case is assigned to one professional rather than another, chance can influence a consequential outcome.
Criminal sentencing makes the stakes especially plain. A basic fairness principle says that similarly situated people convicted of the same offense should not receive dramatically different punishments. Yet judicial discretion was defended as humane: a sentence could reflect the crime, the defendant's character, and individual circumstances. The challenge is that room for judgment can also create unwanted variation.
In 1973, federal judge Marvin Frankel argued that a bank-robbery sentence might range from no prison time to twenty-five years, depending more on the judge's views than on the case. He described comparable counterfeit-check defendants receiving fifteen years and thirty days, and embezzlement defendants receiving 117 days and twenty years. His initial case was forceful but anecdotal. Later controlled studies tested it. A 1974 study of fifty judges given identical pre-sentence reports found heroin recommendations ranging from one to ten years, and bank-robbery recommendations from five to eighteen. In sixteen of twenty cases, judges could not even agree whether imprisonment was appropriate. A 1981 study of 208 federal judges found unanimous agreement to impose prison in only three of sixteen hypothetical cases.
These experiments probably understate real-world noise, because actual cases contain more information, including irrelevant details. Hunger, breaks, local football losses, birthdays, and temperature have all been associated with different legal decisions. The point is not that every sentence has one perfect answer. It is that uncertainty does not excuse arbitrary variation. Even in a difficult, value-laden judgment, wide disagreement among competent decision makers can turn legal authority into a lottery, producing unfairness and costly error.
Chapter 2 of 16 · 4 min · Audio & textIn the app
The Judgment Lottery
After showing how sentencing can vary with the judge, the authors turn to insurance, where the same problem hides inside ordinary organizational work. Executives agreed consistency was desirable and some variation inevitable, but did not expect its scale.
Chapter 3 of 16 · 3 min · Audio & textIn the app
One-Off Decisions Still Vary
Noise is easiest to see when many professionals repeatedly handle comparable cases. Their disagreements can be compared.
Chapter 4 of 16 · 4 min · Audio & textIn the app
Judgment Is Human Measurement
To understand noise, first define judgment precisely. In the authors’ technical sense, a judgment assigns a value to a case or event, often on a scale.
Chapter 5 of 16 · 4 min · Audio & textIn the app
Error Has Two Sources
Judgment is a form of human measurement, so its errors can be analyzed. In a predictive judgment, the target is a true value, such as a future market share or diagnosis.
Chapter 6 of 16 · 4 min · Audio & textIn the app
The Layers of Variability
Earlier, the book separated noise from bias: bias pushes judgments in a common direction, while noise makes them disagree. That disagreement has structure.
Chapter 7 of 16 · 4 min · Audio & textIn the app
Groups Can Amplify Noise
The variability of judgment does not end with the individual. A person's answer can change even when the evidence has not.
Chapter 8 of 16 · 4 min · Audio & textIn the app
When Formulas Beat Experts
Clinical judgment can feel superior to a formula because it seems to notice each person’s complexity. But in prediction, that apparent subtlety can become noise.
Chapter 9 of 16 · 5 min · Audio & textIn the app
The Ceiling of Prediction
Reducing noise is not the same as making the world predictable. A decision can become more consistent and still miss because the future contains facts no judge, formula, or machine could know at the time.
Chapter 10 of 16 · 5 min · Audio & textIn the app
How Minds Manufacture Noise
Noise often begins before a judgment reaches its final form. When evidence is incomplete, the mind uses shortcuts instead of weighing every fact.
Chapter 11 of 16 · 4 min · Audio & textIn the app
Personal Patterns, Practical Hygiene
Earlier, the book separated bias from noise and then divided system noise into layers. The full hierarchy is now complete: total error can contain bias and system noise; system noise contains level and pattern noise; pattern noise contains stable pattern and occasion noise.
Chapter 12 of 16 · 4 min · Audio & textIn the app
Better Judges, Better Processes
Once the book has diagnosed how judgments vary, the practical question is how to prevent error. A single correction cannot do that.
Chapter 13 of 16 · 7 min · Audio & textIn the app
Hygiene Across Real Institutions
Decision hygiene becomes concrete when institutions make high-stakes judgments without a guaranteed answer. Across forensic science, forecasting, and medicine, the authors emphasize sequencing information, independent second opinions, aggregation, judge selection, feedback, calibration, and explicit criteria.
Chapter 14 of 16 · 5 min · Audio & textIn the app
Structuring Organizational Judgment
Organizations often ask managers to judge work that cannot be captured by one clean number. Patient volume, sales, or lines of code may matter, but treating any one measure as complete can distort incentives and ignore context.
Chapter 15 of 16 · 5 min · Audio & textIn the app
Wise Limits on Consistency
Once an institution recognizes that its judgments vary, a tempting response is to demand consistency everywhere. The authors argue for a careful standard: reducing noise is valuable, but zero noise is not an automatic destination.
Chapter 16 of 16 · 4 min · Audio & textIn the app
Auditing and Correcting Judgment
After weighing the costs of consistency, the authors end with a practical discipline: measure before reforming. A noise audit starts by asking how much judgments vary across professionals who perform comparable work.
Chapter 1 of 16 · 4 min · Audio & text: Noise Hides in Plain Sight
Wiseley supports reading and listening to summaries in the app.
Continue in WiseleyAbout Daniel Kahneman, Olivier Sibony & Cass R. Sunstein
Daniel Kahneman
Daniel Kahneman was an Israeli-American psychologist whose work helped connect psychology and economics. His book Thinking, Fast and Slow explores judgment, cognitive biases, and the different ways people make decisions.
Explore more books by Daniel Kahneman
Olivier Sibony
Olivier Sibony is a French author. “Noise” explores the unwanted variability in human judgment and ways to reduce it.
Explore more books by Olivier Sibony
Cass R. Sunstein
Cass R. Sunstein is an American legal scholar who writes about public policy and human decision-making. With economist Richard Thaler, he coauthored Nudge, examining how the design of choices can influence behavior.
Explore more books by Cass R. Sunstein
