Our Noisy Minds

Episode Summary

Executive Summary: In this episode of Hidden Brain, Shankar Vedantam interviews Nobel Prize-winning psychologist Daniel Kahneman about his new book 'Noise: A Flaw in Human Judgment.' Kahneman explains that noise is unwanted variability in human judgments, distinct from bias, and is far more prevalent than people realize. Through examples from insurance underwriting, criminal sentencing, medicine, and asylum cases, he demonstrates how noise leads to costly errors and injustice. The episode explores why noise is often overlooked, the psychological resistance to reducing it, and strategies like decision hygiene and algorithms to combat it.

Main Topics: Definition and Prevalence of Noise (Priority: 5/5): Noise is unwanted variability in judgments where identical cases receive different outcomes. Kahneman shows it's pervasive across domains like insurance, medicine, and criminal justice, often exceeding expectations. Noise vs. Bias (Priority: 5/5): Noise is distinct from bias; bias is systematic error, while noise is random variability. Both contribute to unfairness, but noise is harder to detect and less discussed because it lacks a causal story. Noise in Criminal Justice (Priority: 4/5): Judge Marvin Frankel highlighted sentencing disparities, leading to guidelines that reduced noise but were later struck down. Noise returned, disproportionately harming black defendants. Noise in Medicine (Priority: 4/5): Physicians disagree on diagnoses and treatments, leading to errors. Extraneous factors like time of day affect prescribing rates, contributing to costly mistakes. Psychological Resistance to Reducing Noise (Priority: 4/5): People, especially professionals like judges, resist rules that limit discretion due to naive realism—the belief that their view is objective. They prefer human judgment over algorithms despite evidence. Strategies to Reduce Noise (Priority: 4/5): Techniques include noise audits, averaging independent judgments, using algorithms, and decision hygiene—general procedures to improve judgment quality without targeting specific biases. The Wisdom of Crowds and Self-Averaging (Priority: 3/5): Averaging multiple independent judgments reduces noise. Even averaging one's own estimates over time improves accuracy, as shown in studies by Vul and Pashler.

Key Arguments: Noise is a major source of error that is often overlooked because it is statistically invisible and lacks a causal story, unlike bias. Human judgment is inherently variable, and this variability leads to costly mistakes in insurance, medicine, and criminal justice. Professionals resist constraints on their discretion due to naive realism, believing their judgments are objective and unique. Algorithms can reduce noise and improve accuracy, but people prefer human decision-makers even when they make more errors. Decision hygiene—general procedures to improve judgment—can be more effective than targeting specific biases. Measuring noise through audits is the first step to addressing it, as intuitions about its magnitude are systematically wrong.

Data Points: Expected vs. actual variability in insurance underwriting: 10% expected, 55% actual - Executives at an insurance company estimated variability between underwriters at 10%, but measurement showed 55%. Cost of noise in insurance: Billions of dollars - The overall cost of mistakes from noisy underwriting was estimated in the billions. Asylum judge grant rates in Miami: 88% vs. 5% - One judge granted asylum to 88% of applicants, another to only 5%, showing extreme system noise. Sentencing disparity in federal judges study: Average sentence 7 years, probable difference over 3 years - 208 federal judges evaluated 16 cases; the average sentence was 7 years, but the probable difference between two judges was over 3 years. Reduction in jail population with algorithm: 42% - An algorithm advising judges on bail could reduce jail population by 42% without increasing crime risk. Noise reduction by averaging independent judgments: 50% reduction with 4 judges, 90% with 100 - Averaging four independent judgments reduces noise by half; 100 reduces it by 90%.

Pivotal Quotes: "Wherever there is judgment, there is noise, and there is more of it than you think." — Daniel Kahneman: Kahneman summarizing the core thesis of his book, emphasizing the ubiquity and underestimation of noise. "If it takes a mechanistic science to produce justice, then I think we should seriously consider some mechanistic science." — Daniel Kahneman: Responding to Judge Cabranis's criticism of sentencing guidelines as mechanistic, arguing that consistency is essential for justice. "Noise doesn't lend itself to a causal story. And really, the mind is hungry for causes. And that leads us very naturally to think in terms of biases: that errors must be." — Daniel Kahneman: Explaining why noise is overlooked in favor of bias—noise lacks a narrative, while bias provides a satisfying explanation.

Implications: This episode challenges listeners to recognize noise in their own decisions and organizations. It suggests that reducing noise through measurement, aggregation, and algorithms can improve fairness and efficiency, but requires overcoming psychological resistance. The insights are crucial for leaders in justice, medicine, and business.