NOISE IS UNWANTED variation in judgments that should be identical, which leads to inaccurate and unfair decisions. It is all around people all the time, though individuals fail to notice it. To get a sense of how it happens, perform a“noise audit”right now: open your phone’s stopwatch App and practice counting ten seconds. Now, with your eyes closed, count several times, hitting the lap button each time you believe ten seconds have elapsed.
本该一模一样的判断中出现了人们不想要的差异，这种差异便是「噪音」（noise），它导向了不准确和不公允的决策。噪音无时不有，尽管个体并不会注意到它。要了解它是如何产生的，不妨现在就来做一次「噪音审计」：打开你手机上的秒表应用，先练习一番，数个 10 秒。接着闭上眼睛，再数几遍，每次你觉得已经过了 10 秒，就按停计时器。
Your answers weren’t perfect but noisy: slightly above or below the ten-second mark. And if they were consistently wrong in one direction, then there is bias too, which is a different form of error (you counted too quickly or slowly).
你每次计数的结果并不是正正好好，而是有噪音的：不是比 10 秒稍长些，就是略短了点。如果它们始终都往一个方向上错，那就说明还存在偏误，这就是另一种形式的错误了（你数得太快或太慢了）。
The problem of bias in decisions is well known and there are strategies that people can adopt to minimise it. For example, customers may be“anchored”on the first price they are presented with in a transaction, so they learn to consciously discard it before they negotiate. But noise is different precisely because it is Less apparent.“It becomes visible only when we think statistically about an ensemble of similar judgments. Indeed, it then becomes hard to miss,”Daniel Kahneman, Olivier Sibony and Cass Sunstein write in their new book.
决策时会出现偏误的问题众所周知，人们可以采取一些策略来尽量减小它。例如，在交易中，客户可能会「锚定」在对方给出的第一个报价上，因此他们学会了要在谈判之前有意识地无视这个价格。但噪音之所以不同，恰恰是因为它不是那么明显。「只有当我们从统计学的角度综合考量一系列相似的判断时，噪音才变得明显起来。事实上，这之后你想不看到它们都难。」丹尼尔・卡尼曼（Daniel Kahneman）、奥利维尔・西博尼（Olivier Sibony）和卡斯・桑斯坦（Cass Sunstein）在他们的新书中写道。
The divergences are stark. In a courthouse in Miami, one judge would grant refugees asylum in 88% of cases while another would do so 5% of the time. A large study of radiologists found that the false-positive rate ranged from 1% to 64%, meaning that two-thirds of the time, a radiologist said a mammogram showed cancer when it was not cancerous. Doctors are more likely to prescribe opioids at the end of a long day. Judges made harsher decisions leading up to their breaks and on hotter days. An insurance firm’s underwriters assessed premiums that varied by 55%, a difference that was five times greater than its management had imagined.
分歧十分显著。在迈阿密的一家法院，一名法官裁定给予难民庇护的几率是 88%，另一名法官是 5%。一项针对放射科医生的大规模研究发现假阳性率在 1% 至 64% 之间——64% 这个数字意味着当某位放射科医生说乳房 X 光片显示癌症时，有三分之二的机会都不是。医生更有可能在一天的辛苦工作快结束时开出阿片类药物。法官在临近休假和天气较热时会做出更严厉的判决。在一家保险公司，不同承保人评估的保费相差 55%，是公司管理层预期数字的五倍。
Not only do individuals differ with their peers, they often fail to agree with themselves. Wine experts tasting the same samples for a second time scored fewer than one in five identically. Four out of five fingerprint examiners altered their original identification decision when presented with contextual information that should not have been a factor in matching prints. In one medical study, assessing angiograms, physicians disagreed with their earlier judgments more than half the time.
Noise is sometimes good. When different investors size up a trade or book reviewers reach different assessments, the diversity of opinion is beneficial. But more commonly it creates problems. In law noise means unfairness. In business it can be costly.
Yet it can be reduced. The authors’remedies include a“noise audit”to measure the degree of disagreement on the same cases, to quantify the variation that is usually invisible. They also call for better“decision hygiene”such as designating an observer for group decisions, to prevent common biases and noisy judgments. For example, they can ensure that participants in a team reach independent assessments before coming together as a group to aggregate their decisions.
Another solution is to dispense with people altogether. Statistical models, pre-determined rules and algorithms in many cases are more accurate than human judgment. The authors welcome artificial intelligence to make many decisions in society, but acknowledge that people are predisposed to resisting their answers, for lack of the personal, emotional quality in decision-making—even if it leads to inferior, or at least variable, decisions.
The trio speaks with credibility. Mr Kahneman is a Nobel laureate whose ideas on bias in human reasoning have reshaped economics and society; Mr Sunstein is a polymath scholar at Harvard and occasional government official putting his ideas into policy; Mr Sibony is a former McKinsey partner who teaches decision science at a French business school. Yet despite the book’s title, the authors struggled to extract the signal from the noise, so to speak, needing some 400 pages to make their case. A tighter argument would have enhanced the ideas they present.
三位作者的话颇具可信度。卡尼曼是诺贝尔奖获得者，他关于人在推理时会出现偏误的观点重塑了经济和社会；桑斯坦是哈佛大学一名博学多才的学者，间或在政府部门任职，将自己的想法付诸实际政策；西博尼曾是麦肯锡的合伙人，如今在一所法国商学院教授决策科学。不过，尽管该书以噪音为题，作者们自己要从噪音中提取出信号还是有些费力——乃至要花 400 来页来说明白。如果论证能更紧凑些，他们提出的观点会更有力。