Noise
This one is on the production list — the three-minute video isn't out yet. Want it sooner? Tell us: hello@distilshelf.com with "Noise" in the subject.
What is Noise about?
Daniel Kahneman, with Olivier Sibony and Cass Sunstein, examines noise: unwanted variability in judgments that should be identical. Two judges give wildly different sentences for the same crime, underwriters price the same risk differently. Unlike bias, noise is invisible and unmeasured, and the book argues it is often the bigger source of error, offering decision hygiene as the fix.
Key ideas from Noise
Noise Versus Bias
Bias is systematic error in one direction, noise is scatter. An organization can be unbiased on average while individual judgments vary wildly, and error grows with both, so reducing scatter matters as much as fixing bias.
The Noise Audit
Give many professionals the same case and measure disagreement. Firms doing this discover variability several times larger than executives predicted, in insurance pricing, sentencing, forecasting and hiring, error nobody had ever counted.
Decision Hygiene
Fixes include structuring judgments into independent components, aggregating multiple independent estimates, using relative rather than absolute scales, and delaying holistic intuition until the end. Rules and algorithms, being noise free, often beat experts.
Lines worth keeping
- Wherever there is judgment there is noise, and there is far more of it than anyone believes.
- A fair system is not just one without bias, it is one where the verdict does not depend on which judge you happened to draw.
Is Noise worth reading?
Important idea, laborious book. The distinction between noise and bias genuinely changes how you see hiring, medicine and courts, but the argument is made by the third chapter and then restated with heavy jargon for three hundred more pages. Skim it: read the opening, the noise audit examples and the decision hygiene section, and you have the whole thing.
Daniel Kahneman · 2021 · Psychology & Thinking · Distilshelf editorial · Updated August 2026