Noise A Flaw in Human Judgment

While the world is obsessed with fighting bias, Nobel laureate Daniel Kahneman and his co-authors reveal an equally insidious and costly, yet largely invisible, flaw in human judgment: noise. From criminal sentencing and medical diagnoses to hiring decisions and financial forecasts, unwanted variability in judgments that should be identical leads to rampant unfairness, massive economic losses, and catastrophic errors. This book takes you on a deep dive into the nature of noise, distinguishing it from bias, dissecting its psychological origins, and demonstrating its shocking prevalence in both public and private sectors. More than just a diagnosis, "Noise" provides a practical toolkit of "decision hygiene" strategies—simple, preventive measures like structuring decisions, using guidelines, aggregating independent judgments, and taking an "outside view"—that any organization can implement to improve judgment, reduce error, and create a fairer, more consistent world.

What it argues

Noise A Flaw in Human Judgment

Key ideas it contributes

• Structured Decision Process — The practice of decomposing a complex judgment into a predefined set of smaller, independent assessments (mediating assessments) and evaluating each component separately before forming a final, holistic judgment. This includes methods like structured interviews and the Mediating Assessments Protocol.
• Aggregation of Independent Judgments — The process of combining the independent judgments of multiple individuals to arrive at a single, final judgment. Methods include simple averaging, the Delphi method, prediction markets, and the 'wisdom of select crowds'. Independence of initial judgments is critical for effectiveness.
• Use of Rules and Guidelines — The application of formal, pre-established rules or structured guidelines to constrain discretion and guide judgment. Examples range from simple checklists and scoring systems (Apgar score) to mandatory sentencing guidelines and algorithms.
• Use of Relative Scales and Judgments — The practice of making comparative judgments (e.g., ranking options) rather than absolute judgments on a numeric or adjectival scale. This leverages the human ability for fine-grained comparison and creates a shared frame of reference through common anchors or case scales.
• Sequencing of Information — The deliberate management of the order in which information is presented to a judge to prevent premature conclusions and reduce the impact of biasing information and confirmation bias. It involves revealing only necessary information and documenting judgments at each stage.
• Adoption of Outside View — The practice of viewing a specific case as an instance of a broader reference class and using statistical base rates from that class as an anchor for judgment. This counteracts the tendency to focus only on the unique causal narrative of the specific case (the inside view).
• Selection and Training of Judges — The organizational practice of selecting individuals with superior judgment capabilities (e.g., higher intelligence, actively open-minded thinking) and providing training to improve their judgment skills and reduce susceptibility to biases and noise.
• Judge Characteristics — The stable traits of an individual judge, including general mental ability, cognitive style (e.g., actively open-minded thinking), expertise, and personality. Superior characteristics are associated with a lower baseline level of noise and bias in judgments.
• System Noise — The unwanted variability in professional judgments that should ideally be identical. It is random, unpredictable error, composed of level noise (stable differences in judges' average judgments), pattern noise (idiosyncratic judge-case interactions), and occasion noise (within-judge variability).
• Psychological Bias — Predictable, systematic errors in judgment arising from cognitive mechanisms such as heuristics, prejudgments, and the tendency for excessive coherence. When individual psychological biases are not shared or vary in magnitude, they contribute to system noise.
• Judgment Error — The total deviation of a judgment from the 'true' value or optimal outcome. It is mathematically composed of two components: bias (the average error) and noise (the variability of error). Mean Squared Error (MSE) is the standard measure, calculated as Bias² + Noise².
• Objective Ignorance — The irreducible limit on predictive accuracy due to intractable uncertainty about the future and imperfect information. It represents the portion of an outcome that is fundamentally unknowable or unpredictable at the time of judgment, setting a ceiling on how good any judgment can be.
• Judgment Fairness — The degree to which a judgment system treats similarly situated cases and individuals similarly. It is a key normative goal, and is directly undermined by system noise, which creates arbitrary and inconsistent outcomes based on the lottery of which judge is assigned.
• Organizational Performance — The effectiveness and efficiency of an organization in achieving its goals. Judgment error directly harms performance through increased costs, reduced revenues, poor resource allocation, and damage to reputation and credibility.