A Flaw in Human Judgment
Buy book - Noise by Daniel Kahneman, Olivier Sibony and Cass R. Sunstein
What is the subject of the book Noise?
Noise (2021) is an investigation of the chaotic and expensive impact that randomness plays in human decision-making and decision-making processes. By revealing the processes that underpin the functioning of our brains and society, the authors demonstrate that noise — unwelcome unpredictability in decision-making – is both inevitable and elusive. We can, however, reduce the amount of noise in our judgements and make our environment more equitable by using a few sound methods.
Who has read the book Noise?
- Behavioral economists, psychologists, CEOs, and students are all interested in behavioral economics.
- Anyone who is interested in how humans make decisions and how those decisions affect society should read this book.
- Anyone who is concerned with the correctness and fairness of a process.
Who are Daniel Kahneman, Olivier Sibony, and Cass R. Sunstein, and what do they do?
Daniel Kahneman is an economist and psychologist who is best known for his book Thinking, Fast and Slow, which was published in 1995. For his contributions to economics, Kahneman was honored with the Nobel Prize for Economics in 2002 and the Presidential Medal of Freedom in 2013. He is now a professor emeritus at Princeton University, where he taught for many years.
Cass R. Sunstein is a legal researcher and the author or coauthor of many books, including Nudge, which he coauthored with Richard Thaler. He is also the author of various articles and book chapters. Dr. Cass Sunstein served as a senior administrator in President Barack Obama's administration. He is also the creator and head of the Harvard University Program on Behavioral Economics and Public Policy.
He is a fellow at Oxford University, a former senior partner at McKinsey & Company, and the author of You're About to Make a Terrible Mistake! You're About to Make a Terrible Mistake!
What exactly is in it for me? Uncover the mysteries of the fascinating realm of noise.
Consider the following scenario: you are holding a stopwatch in your hand. Start the watch and then stop it after precisely ten seconds, all without glancing at the time. If you repeat this process many times in a row, you will find that reaching 10 seconds on the dot is almost difficult. Sometimes you'll be a bit short, and other times you'll be a little longer. You may be off by a few milliseconds at times. Other times, you're off by a fraction of a second, or even more than that. In any case, you will wind up with a collection of mistakes that have no obvious pattern and no discernible origin as a result of this small experiment. This is an example of noise, or a series of unpredictably bad decisions. And although your mistakes in this small stopwatch experiment seem harmless enough, as you will soon discover, differences in judgment such as these may have much more severe ramifications. Please accept my invitation to the weird realm of noise.
Among the topics covered in these notes are what the weather has to do with your chances of getting into college, why you – and everyone else – are awful at forecasting the future, and how our narrative-seeking minds may cause havoc with our decisions.
Human judgment may be adversely affected by a variety of unrelated and unexpected circumstances.
Consider the following scenario: you are a high school senior, and you and your closest buddy are both self-described intellectual geeks. To gain a better understanding of the random and odd nature of the type of noise we are talking about here, suppose you are a senior in high school. You and your friend have both received consecutive As in school, aced the SATs, and got admissions interviews at the same Ivy League institution. You go in for your interview and everything goes swimmingly, as you would expect. Your strong academic performance impresses the admissions officer, and you go back to your vehicle across campus, the sun on your face and a cool wind on your back, feeling fantastic. Your buddy has an appointment with the same admissions officer the following day, which is convenient for you. In the same way that your interview went, hers went well. However, as soon as she left, the rain clouds that had been gathering all day burst open, unleashing a torrential deluge.
After a few weeks have passed, you and your partner get letters from the admissions office. It turns out that they have rejected you, but have accepted your buddy instead. Your thoughts are whirling around in your head. Why? What is it that she has that you do not? The first and most important message is as follows: Human judgment may be adversely affected by a variety of unrelated and unexpected circumstances. As behavioral scientist Uri Simonsohn wrote in a 2003 article with the provocative title "Clouds Make Nerds Look Good," the weather may have had a role in the outcome of the election. As a result of his research, Simonsohn found that college admissions officials paid greater attention to grades and test scores on cloudier days.
Alternatively, on brighter days, admissions officials are more attentive to nonacademic characteristics, which means that on the day of your interview, the officer may have been more interested in sports and creative ability than in straight As and SAT scores, as opposed to straight As and SAT scores. Alternatively, it is possible that the admissions officer's choice had nothing to do with the weather at all and had everything to do with the interviewers who came before you. That is, maybe, because those kids were excellent prospects, and the admissions officer just did not want to continue on a losing streak.
But hold on a sec. Other non-relevant variables may have also had a role in the choice. Although he had access to air conditioning in the workplace, the admissions officer could have been hungry or frustrated after his local football club suffered a disappointing loss. He might have thought the sun was too hot, despite having air conditioning in the office. Several studies have shown that each of these seemingly insignificant variables may influence the judgments of bank loan officers, baseball umpires, doctors and judges. It is important to note that in all of these situations, a single individual is continually confronted with essentially the same circumstances while making a variety of decisions. This unpredictability is referred to as occasional noise by researchers, and it is one of the main types of noise. However, it is not the only one.
Noise and bias are not the same thing, but bias may result in noise.
Let's try another thought experiment, this time one that takes place during a carnival. More specifically, it is in a shooting arcade.The two of you have just shot numerous metal pellets at paper targets that have been strung up at the far end of the range with BB guns in your hands. You're both bad shooters, but you're doing it in different ways.The misses on your paper target are dispersed across the target. When seen from a distance, it is clear that there is no pattern. Your photographs have a lot of noise in them. Your friend's paper target, on the other hand, tells a different tale. His shots were grouped together, but none of them hit the target. Low and to the left, his photos are a favorite of mine. His expression conveys the impression that the true bullseye is really down there. Alternatively, his rifle's barrel could have been bent.What is the source of his consistent errors?
The most important lesson here is that noise and bias are not the same thing, but bias may cause noise. Bias is the term used to describe when we commit mistakes on a regular basis. The word is often used to express a bias against or against certain groups of individuals in our daily lives. In the area of psychology, the word is often used to refer to cognitive processes that cause us to make erroneous decisions. Consider the phenomenon of conclusion bias, which leads us to bend our judgements in the direction of a desired result, causing us to perceive information in a skewed manner. Consider the example of the Miami immigration court, where it was discovered that the chances of obtaining refuge varied from 5 to 88 percent depending on which of the two judges presided over the case. These two judges' decisions were almost certainly influenced by bias.It goes without saying that this kind of prejudice may have life-altering effects.
The two Miami judges would each create paper targets with bullets that would be scattered all over the place if the asylum judgments were BB pellets in a BB gun. However, if the asylum judgments of the whole Miami courtroom, including the unpredictable choices of other judges, were plotted on a map, the courthouse's paper goal would seem to be a jumbled mess. System noise is a term used to describe the kind of variability that occurs when judgements inside a system are unjustifiably discordant with one another. Remember the obnoxious Ivy League admissions worker who made a lot of noise on occasion? It's possible that this was caused by prejudice as well. But, regardless of whether we're trying to figure out event noise or system noise, we have to inspect our paper target from a distance. In the case when we hold the target too near to our eyes, the noise becomes unnoticeable. Let us now shift our attention to another area that is prone to noise: forecasts.
When making predictions, we are often swayed by what seems correct at the moment.
A judge who decides on bail has a great deal of responsibility. Should she detain a prisoner in jail awaiting trial or should she allow him out on bail? If she makes a mistake and refuses the defendant bail, he would lose his freedom as well as his employment. His family may possibly be forced to leave their house. None of this suffering will have helped to bring about justice. Alternatively, if she grants bail in error, he may flee and, in the worst-case scenario, commit another crime as punishment.When weighing the pros and disadvantages of releasing the defendant, a bail judge draws on her previous experience and the evidence in front of her to anticipate what the defendant will do if freed. Unfortunately, people – and this includes judges – are notoriously bad at generating accurate forecasts in their respective fields. The most important lesson here is that when we make predictions, we are often led wrong by what seems good to us at the time.
In 2018, a team headed by computation and behavior researcher Sendhil Mullainathan developed an algorithm that generated bail decisions in the court of law. They gave it the outcomes of roughly 760,000 real-world bail hearings and discovered that the algorithm would have reduced both the prison population and the crimes committed by released offenders by 24 to 42 percent if it had been implemented at the same time. Another study discovered that a simple formula that takes into account just two variables – the defendant's age and the number of court dates they have skipped - beats human judges in a variety of situations. So, what is it about algorithms and back-of-the-envelope arithmetic that makes professionals with years of training and experience come up so far behind them? The solution is straightforward. Judges are fallible people.
It is our need for closure that drives our attempts to foresee the future. The aim is to solve a mental problem, and when we come up with the solution, we get an internal signal that says, "Yes, that's it!" A pleasant forecast is one that conforms to our worldview, and the strength of this emotional pleasure often blinds us to a fundamental shortcoming of predictions: their inability to foresee the future. We have no way of knowing what we don't know, and what we do know may be incorrect, inaccurate, or deceptive in some way. The rules and algorithms are likewise deficient in knowledge. In fact, it's much worse. They are, however, devoid of internal signals, world views, and emotional rewards that are associated with them. In a nutshell, they outperform humans because they are not affected by noise.
We don't pay attention to noise since it doesn't make for an interesting narrative.
The fact that notes often begin with a narrative may have caught your attention. As a starting point, we create a time and location, an event that includes a character, someone who has a purpose and challenges to accomplish. We do this because the human mind enjoys a good narrative, and knowledge that is useful inside a story is more likely to be retained. Until now, we've looked at noise and some of the ramifications it has for the criminal justice and college admissions processes. As previously said, anywhere individuals make judgements, there is going to be a lot of noise. But, if there is so much noise around us, why haven't we heard more about it? The most important lesson here is that we disregard noise since it does not make for an interesting narrative.
Much of the psychological understanding gained over the last several decades has focused on our strong connection to narratives. We've discovered that the human mind makes sense of the world by inventing tales to explain what it sees and experiences.
For example, psychologists have discovered what is known as the basic attribution mistake, which refers to our tendency to credit or blame individuals for events that are more accurately described by chance. In other words, we see people and storylines all over the place these days. A psychological process known as naïve realism, which is the self-perpetuating assumption that we see reality exactly as it is, helps to strengthen our narratives by excluding potentially problematic counter-narratives when our reality is questioned. The authors refer to this as the valley of the normal, where the unusual is made comprehensible by assigning a reason in retrospect. When something truly unexpected occurs, the mind attempts to place it within what is known as the valley of the normal.
That leads us to the primary point of this discussion: noise is resistant to narrative. Noise is not causal, and it does not conform to our established patterns of comprehension. If there is a narrative at all, it is a frustrating and seemingly pointless one, at least at first glance. It goes unnoticed because there isn't a narrative to fit the loudness. Our responses are either complete obliviousness or conscious editing, or we recognize it as an instance of bias. After all, bias may be effective in a narrative. It has the ability to cause events.
Noise, on the other hand, can only be detected via statistical analysis of data. When faced with the unpredictability of bail rejections, college admissions, asylum hearings, and employment decisions, it is easy to believe that they are influenced by prejudice. Furthermore, as previously stated, prejudice may be a significant element in certain instances. In contrast, when we take a step back and look at these events as a whole, their random and chaotic character becomes evident.
By averaging numerous, independent assessments of a single question, we can reduce the amount of noise we hear.
Before we proceed, let us briefly review everything we have learnt so far, which is as follows: Wherever there is human judgment, there will be noise, and that noise may have frightening and life-changing implications at times, depending on the situation. And thus, the question inevitably arises: what can we do to address the situation? What can we do to reduce the level of noise?
We should start with the autumn of 1906, when Francis Galton, a polymath and distant cousin of renowned evolutionist Charles Darwin, paid a visit to a county fair in Plymouth, to have a better understanding of the issue. While walking around the booths, he stumbled upon a competition in the ox-weighing area. Galton, who was known for being a theorist of intelligence, among other things, listened intently as over 800 people offered their best estimations of the ox's weight, which piqued his interest. No one correctly identified the answer as 1,198 pounds. When the tournament was finished, Galton requested that the organizers give him the tickets so that he could do statistical research on them.
When the estimates were displayed on a graph, they seemed to be all over the place. in varied and unexpected quantities, here and there and everywhere in between. People were making a lot of noise. The meaning of the villagers' estimations, however, revealed something unexpected to Galton when he computed the meaning of the estimates. It was almost flawless, with just a single pound of variance. One of the most important messages here is that we may reduce noise by averaging several independent judgements on a single topic. During his research, Galton discovered a phenomenon known as the wisdom of the crowd effect.Accumulating independent assessments from many judges and then averaging their answers has been proven to be a reliable method of obtaining something near to the truth in hundreds of different situations.
When you ask individuals to estimate the quantity of jelly beans in a jar, the distance between two randomly selected cities, or the temperature a week from now, their responses will be diverse and unpredictable. They'll make a lot of noise. When the responses are averaged out, the noise in one answer balances out the contradicting noise in the other response. The background noise cancels out on its own. The wisdom of the crowd, on the other hand, comes with some important limitations. First and foremost, each judge must be completely independent of the others. In the case of a group, when you pose a question to the whole group, the individuals react equally to the group as they do to the question itself. Furthermore, the wisdom of the crowd is only valid when every person is considering the exact same situation. Asking each individual for a new inquiry will not get you anywhere in life.
Finally, the wisdom of the multitude does not protect against the possibility of prejudice. Whenever a group has a bias, such as a systematic mistake in judgment, the meaning of the group's answers will simply condense and amplify that bias. A hiring committee that is biased against women, for example, would seem much more pronounced when the committee's decisions on female job candidates are aggregated and averaged.
In order to combat noise, you must first make it apparent via the use of a noise audit.
Until this point, we've spent a significant amount of time discussing judges and the seemingly random and sometimes unexplained variety of the punishments they impose. The random nature of this injustice did not escape the attention of Marvin E. Frankel, a United States District Judge in the Northern District of California. At an early stage in his career, Frankel recognized that he had the ability to, for example, sentence a convicted bank robber to up to 25 years in jail or he could select a sentence of one day in prison. Frankel believed that his final decision was influenced by his own personal beliefs, preferences, and prejudices. In 1973, Frankel released a book in which he demonstrated the differences in punishment for offenses that were essentially identical. One example is a small-time check counterfeiter who received 30 days in prison and another who received 15 years for what was basically the same act.
Individual anecdotes, on the other hand, may be rationalized. As a result, Frankel set out to develop a portrait that would be more lasting and methodical. The important lesson here is that in order to combat noise, you must first make it apparent via the use of a noise audit. Frankel led a research team that asked 208 federal judges to sentence criminal offenders to 16 fictitious scenarios in 1981.The results were published in the journal Criminal Justice. Frankel's team presented the situations to each judge separately, after which they plotted the differences in the judges' recommended punishments in each case, according to Frankel. According to the findings of this research and many others like it, statistical evidence has shown that startling variety rather than consistency has been the rule in criminal punishment.
Using auditing techniques, Frankel demonstrated that the amount of noise in any institution where a pool of specialists takes on cases that are significantly similar can be determined. Here's how to go about it. First and foremost, identify your target. Decide how much variation in judgment is acceptable in your situation. What is an acceptable variation in the payment suggested by different claims adjusters who are examining the same flooded basement could be an issue for an insurance executive. After that, assemble your judges, or claims adjusters, depending on the situation, and provide them with scenarios to consider. Make sure you give them a numerical expression for each decision, such as the number of years a person will spend in jail or the amount of money an insurance company will pay out.
Finally, sketch a diagram of the judges' responses in relation to your bullseye. The result is a diagnostic picture of the noise in your organization. Now, what are you going to do about it?
Decisive hygiene may help to decrease noise, and it is similar to normal hygiene in that it requires discipline and preventative measures.
Visualize yourself on an operating table, ready to undergo a surgical procedure. Immediately before you fall asleep under the effects of general anesthesia, the lead surgeon walks over to the sink and starts washing his hands. She lathers up her hands with soap and then runs them under hot water to disinfect them. It is possible that she has stopped an unknown number of viruses from entering your body with just one simple deed. We can do essentially the same thing using noise. We may significantly reduce the amount of noise in our environment by following a set of rules we'll refer to as decision hygiene. There is a clear message here: Decision hygiene may help to decrease noise, and it is just as important as regular cleanliness in that it requires discipline and prevention.
First and foremost, knowing how to stop and think statistically before making any significant choice is the first step toward establishing decision hygiene - the equivalent of the surgeon washing her hands with soap – is essential. In a previous section, we discovered that our narrative-seeking brains create stories out of everything they encounter. We get right in and give the specifics of a case a sense of cause and significance. Despite the fact that this is a natural inclination, it serves as an invitation to chaos. Instead, you should strive to adopt what Kahneman refers to as a "outside perspective," in which you place each instance in the context of a larger body of comparable situations.
Take, for example, the situation when you have a new CEO at work and you are concerned about her ability to be effective. The CEO's education, reputation, and performance history may provide some hints, but it's a jumble of complicated and possibly deceptive information to go through at this point. Instead of starting from scratch, a less noisy method would look at results from comparable circumstances. For example, you may want to find out the average rate of CEO turnover in your sector, or you might want to see how often new CEOs result in increases in stock prices. When developing a solid statistical framework, it is important to avoid jumping to conclusions too quickly.
We all appreciate a decision that feels good, but you want to be sure that it feels right for the correct reasons. Keep that gut feeling you have about the CEO's alma school or that intuition you have about what happened at her last work in a safe place. Instead of this, reserve the emotional reward for a decision that is consistent with a well-founded assessment of what is most probable. Also, on the subject of complicated and coherent judgments, producing one is emotionally pleasant, yet the pleasure may lead to errors in judgment. If at all feasible, split up difficult situations into distinct questions and delegate them to independent arbiters. Connecting the connections between your CEO's tenure and the stock valuation of your business, for example, may be a fascinating mental puzzle, but it may also be completely meaningless. A cautionary story serves as a reminder of one more important aspect of decision hygiene that we should consider.
It is critical that judges support noise reduction in order for it to be effective.
Judge Frankel was successful in the year 1984. Congress passed the Sentence Reform Act, which was quickly followed by the implementation of stringent sentencing guidelines that were based on a study of 10,000 real-world instances. According to the new regulations, judges may only examine the offense and the defendant's previous history when making their decisions. The judge would give a numerical value to each, and the resultant score would determine the range of potential punishments to be imposed. As a consequence, the amount of noise in sentencing has decreased significantly. For example, prior to the Act, a guy convicted of drug trafficking might have faced a range of sentences ranging from one to 10 years, depending solely on which judge happened to be assigned to the case. Following that, the unpredictability was reduced to a time span of a few months.
Judges from all across the nation expressed their dissatisfaction. They'd spent years honing their sense of justice through study and experience, but suddenly, their discretion had been replaced by a rudimentary mathematical problem. The most important lesson to take away from this is that for noise reduction to be effective, judges must be on board. The Sentencing Reform Act was thrown down by the Supreme Court in 2005 due to technicalities in the legislation. An investigation by Harvard law professor Crystal Yang into 4,000 criminal cases that were sentenced following the act's repeal was completed a few years later. The difference between severe punishments and the national average has more than doubled. Personal values have been developed as a foundation for punishment, and this has become the norm. Noise had returned to the scene.
In retrospect, Judge Frankel and his supporters failed to take an important step in their fight for noise reduction. They had failed in their attempt to bring the judges to a consensus on the ultimate objective of the decision. The aim of making a decision should be accuracy rather than personal expression. When compared to literary criticism, competitive sports, filmmaking, or any other area where variety in viewpoint and style fosters richness and development, heterogeneity among specialists evaluating situations that are essentially identical is a concern. A radiology error occurs when two radiologists independently view the same x-ray and get different findings. One of the radiologists was incorrect. To put it another way, before the judges can walk up to the shooting range, they must first agree on the same target.
Once the judges have agreed that accuracy is the most important consideration, the auditors should ask them to participate in the development of the test scenarios. Failure to do so ensures that the audit will be subjected to hostile examination. Following that, the judges will examine the scope and financial impact of the noise. Kahneman, for example, performed an investigation into an insurance firm and found that underwriters were setting rates for clients that were, on average, 55 percent higher than those established by the company. It was necessary for the underwriters to comprehend the significance of minimizing this noise in order for them to realize that losses from over- and underpricing went into the hundreds of millions of dollars.
When implementing decision hygiene, judges must be involved in the formulation of realistic, system-specific norms that strike a compromise between minimizing noise and mitigating other expenses. Examples include the implementation of a "three strikes" law in the United States, which required life imprisonment for individuals convicted of three crimes as a response to increasing crime levels. The regulation decreased noise, but it did so without taking into consideration a defendant's prior criminal history, the gravity of her offense, or her ability to be a good citizen.
Audits, cleanliness, regulations, habits, and prevention are all important considerations. Noise reduction isn't exactly a glamourous line of employment. But, by this point, you've probably realized that noise comes at a high price. It wastes money, promotes injustice, and leads to personal misery for those involved. It erodes trust in institutions like the legal system, health, education, and the workplace. Now that we've discovered it, it's our responsibility to bring it down.
Noise concludes with a final summation.
The fundamental theme of these notes is that random and undesired variability in human judgment exists everywhere, and that we pay a high price for it, whether we recognize it or not. Fortunately, we can decrease noise if we adopt a preventative attitude and follow the principles of noise reduction. Actionable advice: Tap into the collective knowledge that exists within you. Following up on note five, you discovered that averaging several independent assessments on a single question may help to reduce the noise in those judgments and provide you with an incredibly accurate answer. The thing is, if you ask yourself the same question over and over again, you will get the same result. Give it a go. Over the next several days, ask yourself the following question: what percentage of the world's airports are located in the United States of America? When you add together all of your answers, the average will be quite quiet and remarkably close to reality. Warning: there is a spoiler in this answer: it is 32 percent.
Buy book - Noise by Daniel Kahneman, Olivier Sibony and Cass R. Sunstein
Written by BrookPad Team based on Noise by Daniel Kahneman, Olivier Sibony and Cass R. Sunstein