Noise: A Flaw in Human Judgment

“It is not acceptable for similar people, convicted of the same offense, to end up with dramatically different sentences – say, five years in jail for one and probation for another. And yet in many places, something like that happens.”

The authors begin their campaign to convince the reader of the problem of noise with the stark unfairness of people convicted of the same crime receiving radically different punishments. While probation is a warning, a five-year jail sentence takes away the defendant’s freedom for that length of time. This discrepancy in the punishments urges the reader to want to know more about the injustice behind this.

“While few people object to the principle of judicial discretion, almost everyone disapproves of the magnitude of the disparities it produces. 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.”

Here, the authors draw the distinction between judicial discretion, which overall is a good thing as it allows highly trained, intelligent individuals to evaluate a case, and system noise, which is a marker of the disparities in sentencing that make punishments more a reflection of the judge than the defendant and their crimes. By referring to the ethics of people in general, the authors appeal to the reader’s common sense and their willingness to understand that system noise is a problem.

“Measurement, in everyday life as in science, is the act of using an instrument to assign a value on a scale to an object or event. […] Judgment can therefore be described as measurement in which the instrument is a human mind. Implicit in the notion of measurement is the goal of accuracy – to approach truth and minimize error.”

By drawing the analogy between judgment and measurement, a scientific term in most people’s minds, the authors demonstrate the need for objectivity in the decision-making process. Here, as they do later in the text, the authors will align judgment with accuracy rather than personal expression.

“What made you feel you got the judgment right, or at least right enough to be your answer? We suggest this feeling is an internal signal of judgment completion, unrelated to any outside information. Your answer felt right if it seemed to fit comfortably with the evidence.”

Following the challenge of estimating how long the fictitious Michael Gambardi would last in a CEO role, the authors introduce the concept of an “internal signal of judgment completion.” After weighing the evidence, people would reach an answer that intuitively seemed right. The participant would have strung relevant bits of information together until they reached a plausible narrative of cause and effect that would have further given them the impression that they found the right answer.

“Mood has a measurable influence on what you think: what you notice in your environment, what you retrieve from your memory, how you make sense of these signals. But mood has another, more surprising effect: it also changes how you think.”

The authors demonstrate the power of mood in determining thought processes including judgment. The list of cognitive factors that mood affects include what one pays attention to in the external world, what one retrieves from memory, and how one interprets reality. These factors present a persuasive case for the way mood impacts how a person views a particular situation.

“Groups can go in all sorts of directions, depending in part on factors that should be irrelevant. Who speaks first, who speaks last, who speaks with confidence, who is wearing black, who is seated next to whom, who smiles or frowns or gestures at the right moment – all these factors, and many more, affect outcomes.”

Here, the authors list the many variables that can affect group decision making. All are irrelevant to the matter being discussed but can potently sway the argument one way or another. This reminds readers that high-school style politics based on subjective preference and influence continue even in adult arenas. Thus, the potential for noise in group decision0making is huge unless counteractive measures are taken.

“When we see three, ten, or twenty people embracing some conclusion, we might well underestimate the extent to which they are all following their predecessors. We might think that their shared agreement reflects collective wisdom, even if it reflects the initial views of just a few people. Second, informational cascades can lead groups of people in truly terrible directions.”

The authors offer a warning about the tendency of people to copy each other. They therefore advise treating overwhelming consensus with skepticism. The idea of a proposed strategy reflecting the views of a persuasive few undercuts democracy and can even lead people to ignore their individual wisdom in favor of keeping with the group.

“The illusion of validity is found wherever predictive judgments are made, because of a common failure to distinguish between two stages of the prediction task: evaluating cases on the evidence available and predicting actual outcomes.”

The authors demonstrate how the illusion of validity comes to bear wherever people mistake the coherent picture given by the available evidence for a prediction of the future. The authors remind the reader that predicting the future is impossible, and that being overconfident about one’s ability to predict it based on evidence of past performance will lead to error.

“As long as algorithms are not nearly perfect – and, in many domains, objective ignorance dictates that they will never be – human judgment will not be replaced. That is why it must be improved.”

The authors have shown that algorithms can improve on human judgment. However, their imperfection leads humans to not trust them, as humans gather evidence to support their belief that human judgment is superior. The authors thus concede that there is no substitute for improving human judgment.

“The matching prediction can be justified only if reading precocity and college GPA are perfectly correlated, which is clearly not the case. On the other hand, completely ignoring the information about Julie’s reading age would also be a mistake, because her reading age does provide some relevant information. The optimal prediction must lie between these two extremes of perfect knowledge and zero knowledge.”

The authors discuss the lure and pitfalls of matching information to come up with an accurate answer. They demonstrate the importance of evaluating the information people are given and assessing its importance to the question. Thus, reading precocity has only moderate but not necessarily inconsequential impact on college GPA, and before individuals can complete the task of guessing Julie’s GPA, they must deploy outside thinking to determine how much attention they will pay to her reading precocity.

“Multiple, conflicting cues create the ambiguity that defines difficult judgment problems. Ambiguity also explains why complex problems are noisier than simple ones. The rule is simple: if there is more than one way to see anything, people will vary how they see it.”

The authors explain why ambiguity creates more between-person noise, as the more complex an issue, the more perspectives can be taken from it and the more plans of action of how to deal with it can develop. As most of the serious issues that ail society contain conflicting cues, noise-reduction strategies are useful in parsing through these.

“Personal reactions to individual cases can also produce patterns that are stable but highly specific. Consider what led you to pay more attention to some aspects of Julie’s story than to others. Some details of the case may resonate with your life experience.”

The authors define pattern noise as something “stable but highly specific,” meaning that it is akin to a trend in individual human preferences and behavior. Readers can observe this in themselves in the parts of fictitious Julie’s story that they found most memorable and draw correlations with their life experiences. By extension, readers can expect that others, including those making justice and recruitment decisions, will be affected by their own susceptibility to pattern noise.

“Noise is mostly a product not of level differences but of interactions: how different judges deal with particular defendants, how different teachers deal with particular students […] Noise is mostly a by-product of our uniqueness, of our ‘judgment personality.’”

The notion that noise is a product of interactions rather than biases indicates that it occurs in relationships rather than in individuals. This type of pattern noise is harder to detect, as it must be observed over a number of interactions. Just as an individual’s personality is unique, so is their “judgment personality,” which shapes their tendency to decide in a certain way. The authors argue that it is more helpful to attend to a judgment personality than an actual personality when trying to reduce noise.

“Judgments are both less noisy and less biased when those who make them are well trained, are more intelligent and have the right cognitive style. In other words: good judgments depend on what you know, how well you think, and how you think. Good judges tend to be experienced and smart, but they also tend to be actively open-minded and willing to learn from new information.”

The emphasis on how people think as opposed to what they know in judgment is important. It is the application of facts rather than the quantity of them that leads to accurate decision-making. Moreover, the complementary balance of experience to curiosity puts good judges in a position where they can respond to and evaluate new evidence without being wholly swayed by it.

“Checklists have a long history of improving decisions in high-stakes contexts and are particularly well suited to preventing the repetition of past errors.”

Here, the authors draw attention to the fact that the low-tech checklist with its catalog of biases to look out for is a key tool in improving human judgment. On a more basic level than a computer algorithm, it forces decision observers to consider human error objectively.

“On TV shows like CSI: Crime Scene Investigation […] these are latex-glove-wearing, microscope-wielding hard-science types. But Dror realized that examining fingerprints was clearly a matter of judgment. And as a cognitive neuroscientist, he reasoned that wherever there is judgment, there must be noise.”

Dror’s research overturned the premise that the forensic investigation was purely scientific and objective by drawing attention to the prevalence of judgment in the field. Thus, the reassuring illusion of precision in identifying criminals by their fingerprints gives way to a murkier idea of subjective opinion in making sense of the evidence. The presence of noise in an area as crucial as matching fingertips highlights the importance of noise-prevention strategies.

“Tetlock finds that ‘the strongest predictor of rising into the ranks of superforcasters is perpetual beta, the degree to which one is committed to belief updating and self-improvement’.”

The authors cite the Good Decision Project founder Tetlock’s view that the most prominent trait of superforecasters is their ability to seek relevant new information and adjust their position accordingly. This desired degree of flexibility is the opposite of the rigid judgments and positions that characterize much of today’s social and political discourse.

“Such is our ability to create coherence. As we can often find an imaginary pattern in random data or imagine a shape in the contours of a cloud, we are capable of finding logic in perfectly meaningless answers.”

The authors highlight the power of cause and effect narratives in influencing judgment. They demonstrate how humans are capable of making stories from any bits of information no matter how random or relevant to the question. The analogy of finding identifiable shapes in clouds to the process of making sense of information given to employers by interview candidates indicates the high potential for noise and subsequent error in hiring practices.

“Decision hygiene need not be slow and certainly doesn’t need to be bureaucratic. On the contrary, it promotes challenge and debate, not the stifling consensus that characterizes bureaucracies.”

The authors respond to the expected criticism that their proposal of implementing decision hygiene will be time-consuming and stultify progress. They point out that the challenge and debate inherent in such a strategy will be invigorating. Thus, instead of promoting stagnancy, decision hygiene will lessen corruption and help the organization perform better in the future.

“Handwashing does not prevent all diseases. Likewise, decision hygiene will not prevent all mistakes. It will not make every decision brilliant. But like handwashing, it addresses an invisible yet pervasive and damaging problem. Wherever there is judgment, there is noise, and we propose decision hygiene as a tool to reduce it.”

The analogy of decision hygiene and handwashing is a potent one, given the time of writing, which was during the global coronavirus pandemic where people were being urged to wash their hands as a precaution against spreading the virus. Just as handwashing is a hygiene that reduces exposure to harmful germs without preventing them altogether, decision hygiene acts analogously where judgment is concerned, as some irrelevant factors can be identified and eliminated. The effect is thus cumulative and statistical rather than related to a single action or decision.

“Many judgments, including long-term forecasts and answers to fictitious questions, are unverifiable. The quality of such judgments can be assessed only by the quality of the thought process that produces them.”

Where individuals cannot be certain of the future and yet must make decisions about how best to protect or prepare for it, the quality of one’s thinking and the process by which they organize thoughts is imperative. The authors therefore advocate for trying to improve the thought processes that go into judgment-making rather than trying to haphazardly predict outcomes.

“Matters of judgment are characterized by an expectation of bounded disagreement. They occupy a space between matters of computation, where disagreement is not allowed, and matters of taste, where there is little expectation of agreement except in extreme cases.”

The fact that judgment occupies the nebulous space between the objectivity of facts and the caprice of personal taste is a key reason for why it is so complex. It is somewhere between the objectivity of mathematics and the subjectiveness of a wine-tasting contest. The polarity of these two extremes makes a strong case for setting guidelines about how judgments that affect people’s life changes should be made.

“The goal of judgment is accuracy, not individual expression.”

This brief, conclusive statement highlights the authors’ core beliefs where judgment is concerned. They allude to an objective truth that judges are obligated to aim for, rather than the notion of a judge’s subjective preferences influencing outcomes and thereby perpetrating unfairness. These latter things become a source of noise.