Insurance adjusters, judges, doctors, and forecasters of all types have to take incomplete and conflicting information and come up with a decision, often quickly. In many cases they have a lot of freedom in the scope of their decision. A judge can impose a 10 year sentence to one criminal while letting another off for time served, all based on a flawed and noisy system of criteria. Why do we put up with so much variability when the stakes are literally life and death? How can two doctors, lawyers, or accountants come up with totally different answers to complex or ambiguous questions?
There are two blind spots that we humans have when it comes to taking in information and making decisions. One is bias, which psychologists have studied for decades. We are typically biased to favor those who are like us, and our belief systems are ridden with dozens of faulty assumptions that corrupt our perceptions and decisions. But the second blind spot is something that until now had not been written about much- noise.
Noise is distinct from bias because it is all over the place, while bias tends to pull perceptions in one direction predictably. Noise can depend on the time of day a decision takes place, the weather, your mood, things that came before- either good or bad events, or just blind luck. It can make what should be a narrow range of outcomes become a wide spectrum of results that make no sense given the training of those who are tasked with making the decision.
Daniel Kahneman is a widely respected behavioral economist, Princeton professor, Nobel prize winner, and prolific writer. He, along with his co-writers Cass Sunstein and Olivier Sibony have put out a definitive story about the problems of noise in decision-making. They first noticed the problem with a study of insurance companies. They were trying to figure out why some adjusters and underwriters were so far apart in their estimates and findings. For the same disaster, sometimes one homeowner can get tens of thousands of dollars more in money than a neighbor with the same insurance coverage who suffer the same disaster. It shouldn't vary that much. But when Kahneman and his cohorts did what they call a "noise audit", the predicted 10% variability turned out to be more like a whopping 55%, meaning the judgements of the adjusters were all over the place.
Sometimes noise is a good thing, especially in the world of the arts, where tastes vary and one person's favorite song is another's waste of time. But in other areas, like medicine, business and law, you want some predictability that the same criteria will be used consistently.
This discovery of noise is not a good omen for the coming artificial intelligence revolution and how it will effect jobs. Kahneman claims that simple rules and algorithms beat human judgement most of the time. Humans are flawed not only in their numerous biases, but with this problem of noise that crops up when we get distracted, hungry, tired, or moody. People tend to distrust algorithms and computers, which is why so many high-stakes decisions are still left to the humans, but questions need to be asked about who does an all-around better job.
Other humans, and groups, can add noise to our decisions, and Kahneman presents research on how we are directly influenced by what others say and do, especially if they do it loudly or first, creating a cascade of information that amplifies minor differences. The irony is that there is such a thing as "wisdom of the crowd". Given a problem, large groups of people can collectively zero in on an answer, but only if they do it in isolation and then their results are aggregated later. If people are doing a task together, they look to each other and not to the problem.
The authors zoom in on five areas where noise is the worst.
1- Medicine, where fatigue, stress, and human error can combine with disastrous results if noise prevents accurate diagnoses. Psychiatry is especially bad at coming up with noise-free decisions and treatments.
2- Business, where the hiring decision is beset with noisy factors like poor interview questions, assumptions made on first impressions, as well as bias in hiring people who look a certain way. Performance reviews for current employees are also loaded with noisy assumptions and often off the mark.
3- Law, especially judges, who are given wide areas of discretion when pronouncing judgements on people. The problem of noise and bias in judging has been known about for a long time, and attempts at setting stricter guidelines for judges have come and gone.
4- Forecasters like pollsters and economic analysts who try to predict the future and identify trends. Sometimes the most famous experts are the ones who are wrong the most time with their predictions. Re-evaluating predictions repeatedly, and aggregating independent ones, can help reduce the likelihood of huge errors.
5- Forensic scientists. Fingerprint identification is much more noisy than most of us assume. Fingerprints left at the scene of a crime are usually in much rougher condition than those left on a fingerprint card. Making a positive ID often requires independent attempts by different experts to avoid the possibility of noise.
Noise causes more errors than bias according to the authors. Obviously one solution is to pick better judges- ones who are experts in the field and who approach problems openly and with the big picture in mind. They can't be afraid of being wrong and must be willing to re-evaluate things from time to time.
The other half of the solutions is what they call decision hygiene. In order to keep noise from messing up our decisions we need to practice the following:
- Remember that the goal is accuracy, not self-expression
- Think statistically and take the outside view
- Break judgements down into several independent tasks
- Resist your first impressions and intuitions
- Get independent judgements from multiple judges, and then consider aggregating them.
- Favor relative judgements to absolute judgements. Place on a scale of 1-10.
The audience for this book is people who use judgement for a living like doctors, judges, pollsters, and forensic investigators. But for the average person, important judgements like who to marry, where to work, and how to live can be clouded with both bias and noise as well Knowing the different factors that are working against us helps us notice them and adapt to them.
There will never be a noiseless world, nor should there be. No judgement can be certain or 100% right. But important judgements, like whether to deny bail, how much to pay out in insurance claims, and whether someone has cancer or not, have profound effect when they are made. This book presents some important research that will hopefully clean up the messes that pollute our brains.
At some point in all of our lives, we will be depending on another human to evaluate something about us and render a judgment. Could be a doctor or just about anybody who could either open or close doors for us. Books like this one help add to the discussion about how to have the fairest, cleanest, and most accurate judgements that we can make. Avoiding the traps of both bias and noise takes us closer to our true selves.
