Chapter 1: Introduction: Two Kinds of Error
Key concepts: Introduction: Two Kinds of Error
1. Introduction: Two Kinds of Error
The Shooting Range Analogy: Distinguishing Bias from Noise
- Bias is systematic, predictable error (like a bent gunsight causing consistent misses).
- Noise is random, unwanted variability in judgments (like shots scattering widely).
- Some systems suffer from both bias and noise, compounding error.
- The analogy translates directly to human judgment: bias is directional inaccuracy, noise is inconsistency.
Measuring Noise Without Knowing the Truth
- Noise can be identified by examining variability alone, even when the correct answer is unknown.
- Disagreement among experts (e.g., doctors, judges) reveals noise regardless of who is right.
- This 'view from behind the target' makes noise a measurable problem in uncertain real-world decisions.
Evidence of Pervasive Noise in Critical Domains
- Medical diagnoses vary widely between doctors for the same patient.
- Judicial systems show extreme inconsistency (e.g., asylum grant rates from 5% to 88%).
- Professional forecasts, personnel decisions, and forensic science exhibit high variability.
- This noise creates unfairness ('refugee roulette') and economic waste.
Book Structure and Purpose
- Aims to elevate noise in public awareness as a distinct and damaging error.
- Will investigate noise in specific domains (criminal justice, insurance) via noise audits.
- Explores the psychology of why noise is overlooked and the advantages of algorithms.
- Provides practical noise-reduction techniques under the framework of 'decision hygiene'.
- Considers the appropriate level of noise, acknowledging it cannot always be eliminated entirely.
