202301061108 Status: #idea #🧠 Tags: #psychology #decision-making #humans # Noise vs Bias ## Overview Errors in decision-making are typically the result of two factors: *bias* and *noise*. Bias describes a predicable and consistent tendency to favour one direction or another, typically based on a noticable preference: a manager might be biased towards hiring men over women; a judge might express bias by punishing black defendants more harshly. Noise describes a less predictable (and less conisistent) variation in judgements, especially in those judgements we typically deem to be consistent or mechanical. It's the inconsistency we find when we expect consistency, and can't explain easily with causal stories. While bias has been studied extensively and discussed at length, noise is more of a silent killer. This is because bias better fits with our preference for finding causal narratives, whereas noise requires a more statistical mindset; considering any given judgement as a recurring judgement that's been made only once. Any given error in judgement typically contains both noise and bias. In all decisions that are supposed to be mechanical (ie. consistent) basic algorithms will always outperform human judges, because they will reduce noise. In order to control for noise, an open-minded, statistical approach is necessary — along with a pretty big dose of humility. This includes conducting *noise audits* and sticking to a [[Bias Observation Checklist]]. The authors suggest that the best, most human approach to decision making is to first decompose and then sequence any given decision in the most statistical, objective way possible, obtain a pluraity of perspectives from multiple judges (without having their opinions contaminated by one another), average the answer out to arrive at an unbiased mean, and then, if necessary, make a judgement call from there. ## Chapters ### Introduction: Two Kinds of Error Outlines the difference between bias and noise, and decomposes any error into its two component pieces using MSE. #### Part I: Finding Noise There are unexpected levels of noise in our daily judgements, particularlly focussing on system-level noise in institutions where we don't expect it, like organisations and criminal justice. The core idea here is *wherever there is a decision there is noise, and more of it than you think* #### Part II: Your Mind Is a Measuring Instrument We mistake our own minds for an objective decision-making tool, but that's not how it's designed. Our minds are inconsistent and fallible, heavily influenced by our context — moods, how much sleep we've gotten, etc. By way of illustration, the authors ask you to consistently measure something as simple as 10 seconds. The variation betwen each measurement is "noise". We're also heavily influeced by the decisions of others. While the "wisdowm of crowds" will lead to strikingly accurate judgements if each individual judgement is made independently and a mean (average) judgement is taken, individuals will be easily led by the opinions of others. Therefore if we want to benefit from diverse opinions (and we do) we should work hard to make sure decisions are not contaminated by groupthink. #### Part III: Noise in Predictive Judgments Almost all decisions are made on incomplete information. Some facts and data are unknowable (this is called *objective ignorance*) and some facts are simply unknown by the judge. However, we're often confident in our own predictions. This is because we mistake objectivity with a subjective feeling of conidence that comes from satisfying a need for a coherent, causal narrative. If an outcome fits within the *valley of the normal* — i.e. doesn't defy our expectations of what's likely or possible — we have no cause to second-guess ourselves. That is to say, if it seems plausible, we'll just assume we're right. This is particularlly prevalent in *hindsight bias*, where we're able to contruct narratives of causal relationships between events easily, as if any given outcome were inevitable — when, in reality, any given outcome could have gone a number of other ways that would have been equally as plausible. The sum of any amount of these alternative outcome would have aggreagated into an equally plausible counterfactual that would have also fallen into the *valley of the normal*; this is things that seem obvious in hindsight are still hard to predict. We typically only tend to look harder for causal explanations when a given situation defies these expectations of normality, like when Trump was eleceted. #### Part IV: How Noise Happens Goes into a lot of the work covered by [[Thinking Fast and Slow]]. The most interesting point on [[Heuristics]] is the idea of substitution: we tend to substitute hard questions or decisions for ones that are simpler. This gives us the feeling of comfort that comes from coherence. #### Part V: Improving Judgments Some judges are better than others; typically these are people who are highly educated and knowledgeable, but more importantly good judges have a particular approach which allows them to take an *outside view* and constantly update their opinions based on new data as it comes in. They're open-minded and more statistically oriented. The authors make a destinction between genuinely good judges (and *superforecasters*) and *respect experts* who are people that we trust based on their status. A good indication of intelligence is GMA (because IQ is no longer in vogue). Though there are a number of issues with how G is measured, and we need to be careful with its implications, it's impossible to dismiss the correlation between high levels of G and success in judgements. #### Part VI: Optimal Noise Is there an inhumanity to reducing noise? While diverse tastes and opinions are are what makes life interesting (and what sparks progress) we need to be careful where and when we alllow noise to get in the way of decisions. The general idea here is to be conscious of when we're leaning into tastes and perspectives (intuitions and judgements) and when we're expecting a mechanical and consistent outcome. ## Conclusion: As we use the term, judgment should not be confused with “thinking.” It is a much narrower concept: judgment is a form of measurement in which the instrument is a human mind. Similarly, the best way to think about singular judgments is to treat them as recurrent judgments that are made only once. That is why decision hygiene should improve them, too. --- Many people earn a living by making professional judgments, and everyone is affected by such judgments in important ways. Professional judges, as we call them here, include football coaches and cardiologists, lawyers and engineers, Hollywood executives and insurance underwriters, and many more. — location: [5268](kindle://book?action=open&asin=B08LCZFJZ2&location=5268) ^ref-9546 --- Some judgments are predictive, and some predictive judgments are verifiable; we will eventually know whether they were accurate. This is generally the case for short-term forecasts of outcomes such as the effects of a medication, the course of a pandemic, or the results of an election. But 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. Furthermore, many judgments are not predictive but… — location: [5272](kindle://book?action=open&asin=B08LCZFJZ2&location=5272) ^ref-4137 --- people who make judgments behave as if a true value exists, regardless of whether it does. They think and act as if there were an invisible bull’s-eye at which to aim, one that they and others should not miss by much. The phrase judgment call implies both the possibility of disagreement and the expectation that it will be… — location: [5277](kindle://book?action=open&asin=B08LCZFJZ2&location=5277) ^ref-12988 --- We say that bias exists when most errors in a set of judgments are in the same direction. Bias is the average error, as, for example, when a team of shooters consistently hits below and to the left of the target; when executives are too optimistic about sales, year after year; or when a company keeps reinvesting money in failing projects that it should write off. Eliminating bias from a set of judgments will not eliminate all error. The errors that remain when bias is removed are not shared. They are the unwanted divergence of judgments, the unreliability of the measuring instrument we apply to reality. They are noise. Noise is variability in judgments that should be identical. We use the term system noise for the noise observed in… — location: [5282](kindle://book?action=open&asin=B08LCZFJZ2&location=5282) ^ref-2692 --- The mean of squared errors (MSE) has been the standard of accuracy in scientific measurement for two hundred years. The main features of MSE are that it yields the sample mean as an unbiased estimate of the population mean, treats positive and… — location: [5290](kindle://book?action=open&asin=B08LCZFJZ2&location=5290) ^ref-54057 --- MSE does not reflect the real costs of judgment errors, which are often asymmetric. However, professional decisions always require accurate predictions. For a city facing a hurricane, the costs of under- and overestimating the threat are clearly not the same, but you would not want these costs to influence the meteorologists’ forecast of the storm’s speed and trajectory. MSE is the… — location: [5292](kindle://book?action=open&asin=B08LCZFJZ2&location=5292) ^ref-14859 --- Bias and noise make equal contributions to overall error (MSE) when the mean of errors (the bias) is equal to the standard deviations of errors (the noise). When the distribution of judgments is normal (the standard bell-shaped curve), the effects of bias and noise are equal when 84% of judgments are above (or below) the true value. This is a substantial bias, which will often be detectable in a professional context. When the bias is smaller than one standard deviation, noise is the bigger source of overall error. — location: [5308](kindle://book?action=open&asin=B08LCZFJZ2&location=5308) ^ref-41902 --- The large role of noise in error contradicts a commonly held belief that random errors do not matter, because they “cancel out.” This belief is wrong. If multiple shots are scattered around the target, it is unhelpful to say that, on average, they hit the bull’s-eye. — location: [5319](kindle://book?action=open&asin=B08LCZFJZ2&location=5319) ^ref-36670 --- System noise can be broken down into level noise and pattern noise. Some judges are generally more severe than others, and others are more lenient; some forecasters are generally bullish and others bearish about market prospects; some doctors prescribe more antibiotics than others do. — location: [5326](kindle://book?action=open&asin=B08LCZFJZ2&location=5326) ^ref-10375 --- Level noise is the variability of the average judgments made by different individuals. The ambiguity of judgment scales is one of the sources of level noise. Words such as likely or numbers (e.g., “4 on a scale of 0 to 6”) mean different things to different people. — location: [5328](kindle://book?action=open&asin=B08LCZFJZ2&location=5328) ^ref-21886 --- System noise includes another, generally larger component. Regardless of the average level of their judgments, two judges may differ in their views of which crimes deserve the harsher sentences. Their sentencing decisions will produce a different ranking of cases. We call this variability pattern noise (the technical term is statistical interaction). — location: [5331](kindle://book?action=open&asin=B08LCZFJZ2&location=5331) ^ref-24712 --- The main source of pattern noise is stable: it is the difference in the personal, idiosyncratic responses of judges to the same case. — location: [5333](kindle://book?action=open&asin=B08LCZFJZ2&location=5333) ^ref-23512 --- This stable pattern noise reflects the uniqueness of judges: their response to cases is as individual as their personality. The subtle differences among people are often enjoyable and interesting, but the differences become problematic when professionals operate within a system that assumes consistency. — location: [5339](kindle://book?action=open&asin=B08LCZFJZ2&location=5339) ^ref-44035 --- Pattern noise also has a transient component, called occasion noise. We detect this kind of noise if a radiologist assigns different diagnoses to the same image on different days or if a fingerprint examiner identifies two prints as a match on one occasion but not on another. — location: [5343](kindle://book?action=open&asin=B08LCZFJZ2&location=5343) ^ref-8913 --- The judges’ cognitive flaws are not the only cause of errors in predictive judgments. Objective ignorance often plays a larger role. Some facts are actually unknowable—how many grandchildren a baby born yesterday will have seventy years from now, or the number of a winning lottery ticket in a drawing to be held next year. Others are perhaps knowable but are not known to the judge. People’s exaggerated confidence in their predictive judgment underestimates their objective ignorance as well as their biases. — location: [5348](kindle://book?action=open&asin=B08LCZFJZ2&location=5348) ^ref-50082 --- There is a limit to the accuracy of our predictions, and this limit is often quite low. Nevertheless, we are generally comfortable with our judgments. What gives us this satisfying confidence is an internal signal, a self-generated reward for fitting the facts and the judgment into a coherent story. — location: [5352](kindle://book?action=open&asin=B08LCZFJZ2&location=5352) ^ref-39633 --- Psychological biases are, of course, a source of systematic error, or statistical bias. Less obviously, they are also a source of noise. When biases are not shared by all judges, when they are present to different degrees, and when their effects depend on extraneous circumstances, psychological biases produce noise. For instance, if half the managers who make hiring decisions are biased against women and half are biased in their favor, there will be no overall bias, but system noise will cause many hiring errors. Another example is the disproportionate effect of first impressions. — location: [5360](kindle://book?action=open&asin=B08LCZFJZ2&location=5360) ^ref-2653 --- Large individual differences emerge when a judgment requires the weighting of multiple, conflicting cues. Looking at the same candidate, some recruiters will give more weight to evidence of brilliance or charisma; others will be more influenced by concerns about diligence or calm under pressure. When cues are inconsistent and do not fit a coherent story, different people will inevitably give more weight to certain cues and ignore others. Pattern noise will result. — location: [5371](kindle://book?action=open&asin=B08LCZFJZ2&location=5371) ^ref-1082 --- Whenever something goes wrong, we look for a cause—and often find it. In many cases, the cause will appear to be a bias. Bias has a kind of explanatory charisma, which noise lacks. If we try to explain, in hindsight, why a particular decision was wrong, we will easily find bias and never find noise. Only a statistical view of the world enables us to see noise, but that view does not come naturally—we prefer causal stories. — location: [5379](kindle://book?action=open&asin=B08LCZFJZ2&location=5379) ^ref-23668 --- In most fields, a judgment may never be evaluated against a true value and will at most be subjected to vetting by another professional who is considered a respect-expert. Only occasionally will professionals be faced with a surprising disagreement, and when that happens, they will generally find reasons to view it as an isolated case. — location: [5387](kindle://book?action=open&asin=B08LCZFJZ2&location=5387) ^ref-57013 --- There is reason to believe that some people make better judgments than others do. Task-specific skill, intelligence, and a certain cognitive style—best described as being actively open-minded—characterize the best judges. — location: [5392](kindle://book?action=open&asin=B08LCZFJZ2&location=5392) ^ref-50864 --- One strategy for error reduction is debiasing. Typically, people attempt to remove bias from their judgments either by correcting judgments after the fact or by taming biases before they affect judgments. We propose a third option, which is particularly applicable to decisions made in a group setting: detect biases in real time, by designating a decision observer to identify signs of bias (see appendix B). — location: [5396](kindle://book?action=open&asin=B08LCZFJZ2&location=5396) ^ref-27188 --- Our main suggestion for reducing noise in judgment is decision hygiene. — location: [5399](kindle://book?action=open&asin=B08LCZFJZ2&location=5399) ^ref-26825 --- Decision hygiene is as unglamorous as its name and certainly less exciting than a victorious fight against predictable biases. There may be no glory in preventing an unidentified harm, but it is very much worth doing. — location: [5402](kindle://book?action=open&asin=B08LCZFJZ2&location=5402) ^ref-374 --- The goal of judgment is accuracy, not individual expression. This statement is our candidate for the first principle of decision hygiene in judgment. — location: [5408](kindle://book?action=open&asin=B08LCZFJZ2&location=5408) ^ref-28314 --- To be clear, personal values, individuality, and creativity are needed, even essential, in many phases of thinking and decision making, including the choice of goals, the formulation of novel ways to approach a problem, and the generation of options. But when it comes to making a judgment about these options, expressions of individuality are a source of noise. — location: [5412](kindle://book?action=open&asin=B08LCZFJZ2&location=5412) ^ref-58871 --- Think statistically, and take the outside view of the case. We say that a judge takes the outside view of a case when she considers it as a member of a reference class of similar cases rather than as a unique problem. This approach diverges from the default mode of thinking, which focuses firmly on the case at hand and embeds it in a causal story. — location: [5421](kindle://book?action=open&asin=B08LCZFJZ2&location=5421) ^ref-28709 --- Structure judgments into several independent tasks. This divide-and-conquer principle is made necessary by the psychological mechanism we have described as excessive coherence, which causes people to distort or ignore information that does not fit a preexisting or emerging story. Overall accuracy suffers when impressions of distinct aspects of a case contaminate each other. — location: [5430](kindle://book?action=open&asin=B08LCZFJZ2&location=5430) ^ref-53964 --- Resist premature intuitions. We have described the internal signal of judgment completion that gives decision makers confidence in their judgment. — location: [5439](kindle://book?action=open&asin=B08LCZFJZ2&location=5439) ^ref-63845 --- This principle inspires our recommendation to sequence the information: professionals who make judgments should not be given information that they don’t need and that could bias them, even if that information is accurate. — location: [5444](kindle://book?action=open&asin=B08LCZFJZ2&location=5444) ^ref-27843 --- Obtain independent judgments from multiple judges, then consider aggregating those judgments. The requirement of independence is routinely violated in the procedures of organizations, notably in meetings in which participants’ opinions are shaped by those of others. — location: [5448](kindle://book?action=open&asin=B08LCZFJZ2&location=5448) ^ref-30610 --- Favor relative judgments and relative scales. Relative judgments are less noisy than absolute ones, because our ability to categorize objects on a scale is limited, while our ability to make pairwise comparisons is much better. — location: [5455](kindle://book?action=open&asin=B08LCZFJZ2&location=5455) ^ref-50121 --- # References [[Kahneman_et_al-Noise]]