Skip to content

Team Performance Science Guide · Books · 19 of 22

Noise: A Flaw in Human Judgment

Daniel Kahneman, Olivier Sibony & Cass R. Sunstein · 2021

Judgment & decision science·Stance: Ally

Ally — the why-measure-not-opine case

File: ~/Desktop/Performance Books/Noise A Flaw in Human Judgment ...pdf · Tradition: judgment & decision science Stance: Ally — the why-measure-not-opine case (backs Part VII + dossier B; pairs with the idiosyncratic rater effect)

Core claim

"Wherever there is judgment, there is noise — and more of it than you think." Noise = system noise, unwanted variability in judgments that should be identical. It's distinct from bias, often the bigger problem, and almost always under-recognized — and it can be reduced.

Grounded

  • Key theme (verbatim): "wherever there is judgment, there is noise—and more of it than you think."
  • "System noise, that is, unwanted variability in judgments that should ideally be identical, can create rampant injustice, high economic costs, and errors of many kinds."
  • Part II title: "Your Mind Is a Measuring Instrument" — judgments measured for accuracy and error; occasion noise (same case, same person, different occasions, different answer).
  • "Bias is the star of the show. Noise is a bit player, usually offstage" — yet often the larger error.

Contrast / use

This generalizes the idiosyncratic rater effect (Nine Lies, dossier B): rating a person is judgment, so it's noisy — different raters, and even the same rater on different days, disagree on identical cases. The orthodox fix (more training, better scales) addresses bias, not noise. Performix's answer is the structural one Kahneman et al. endorse: replace unaided judgment with measurement (aggregate, decompose, state uncertainty) — "your mind is a measuring instrument," so use a calibrated one. Strong support for psychometric-first (Part VII) and the "manager's gut isn't the cure" seam (#11).

Where it plugs in

Dossier §B (rating noise); Part VII Principle 2 (latent estimation vs noisy judgment); Part I·b (worse-than-arbitrary = bias + noise). Acquired ✓.

Grounds these Guide parts

Library record

No canonical library id minted yet. The PeopleAnalyst library registry stays the identity layer; this profile will carry the id once the work is ingested.

← All books