Book Review — Noise: A Flaw in Human Judgment by Daniel Kahneman, Oliver Sibony and Cass R Sunstein

By: |
October 03, 2021 1:30 AM

A case to cancel out human as well as AI biases while taking decisions

noise book reviewThe point that the authors make is that while laws or rules are objective, human interpretation makes them subjective.

Noise by Nobel laureate Daniel Kahneman, Sibony and Sunstein is an amazing book that takes us into the labyrinths of how we respond variably under different conditions that results in inconsistencies while making any judgments. This being the case, various people judging the same issue will naturally have differential verdicts. And when the verdict involves judging a crime, it goes without saying the result can mean life or death.

Starting with the subject of how different judges look at crimes and give different sentences—some lenient while others are unforgiving—several lives may be pushed to the corner on account of being in the wrong court. It is hence not surprising that various judges are known to be very intransigent when it comes to specific kinds of crimes and would tend to be ‘harsher’ in relative terms than others. But it is not just the case of crime where such ‘noise’ comes in. Even in normal businesses, like insurance, the official who surveys the damages may look at things differently and accordingly recommend compensation. This can affect the outcome of the claim.

The point that the authors make is that while laws or rules are objective, human interpretation makes them subjective. This happens everywhere and often we don’t stop to think. Even in innocuous events like beauty contests, the judges’ views could be clouded. In fact, Keynes had said that judges normally don’t select who they think are the most beautiful, but those who they think others believe are the top rated. Therefore, the issue of subjectivity plays in a big way.

Another area where the authors discuss a lot is the field of medicine. Here, the reader will be able to connect quite easily because rarely do two doctors give the same treatment. And we often talk of how some doctors always recommend several tests or inundate the patient with a plethora of pills while others are less demanding and believe in minimalism. Here, too, it is a case of subjective judgment.

How do we reduce noise? Even setting rules that guide judgments will not eliminate the same, as interpretation is important and it goes back then to the human being who takes a call. Using algorithms is a way out, say the authors, where several court judgements can be evaluated by such micro formulas that reduce individual biases.

In the West, often if the convict is poor and coloured and resides in certain localities, there is an inherent bias that manifests in the final judgment call made. It can be argued that even when algorithms are formulated, it would be essential to ensure that they should go through several rounds to ensure there are no biases that come in.

But there are criticisms too where algorithms could make several silly mistakes, as they may fail to include some answers to questions never asked when being formulated. Sometimes it is recommended that there should be multiple judges for, say, an essay competition. This may work well, though the background of the judges is again necessary. Further, having multiple people give judgments may become unwieldy and inefficient. Therefore, this, too, may not be the best way out.

Will having rules work? Here, too, the authors show how approaching two different insurance offices for a quote can give different opinions even though the company is the same.

If one traces what happens to a cheque which is deposited, the process is well defined, and rules can be applied with no deviation. But not so when there is any element of judgment that can be applied by any human being.

Another option provided by the authors is that the judges should look at the cases statistically, which means examine what were outcomes of similar cases in the past across other courts and not just the specific ones they are evaluating. Here, too, there can be problems of biases in earlier cases that can just be carried through in the new ones.

Another way out is to break the case into several parts and judge them independently. This, however, may not really be possible in reality.

The authors argue that even though the cost of removing noise can be high, it may be worth the effort as it leads to a better solution. This can be justified when it comes to life or death sentences, but if the issue is minor, may impose a monetary and time cost on society. Here we can think of interviews for jobs where some companies put a person through a series of interactions that can be six-10 layers. The idea is not so much to pressure the candidate but to ensure that these subjective biases are adjusted for. There can still be mistakes made, but such layering allows for better judgment.

In short, one cannot really avoid noise or bias in life and several decisions taken are based on subjective judgments. Human beings are unreliable decision makers. These can at the extreme level affect the lives of convicts and hence need to be reduced. But while it can be lowered, it is impossible to eliminate the same and that is the crux of the issue of noise.

Madan Sabnavis is chief economist, CARE Ratings

Noise: A Flaw in Human Judgment
Daniel Kahneman, Oliver Sibony and Cass R Sunstein
William Collins
Pp 454, £16.99

Get live Stock Prices from BSE, NSE, US Market and latest NAV, portfolio of Mutual Funds, Check out latest IPO News, Best Performing IPOs, calculate your tax by Income Tax Calculator, know market’s Top Gainers, Top Losers & Best Equity Funds. Like us on Facebook and follow us on Twitter.

Financial Express is now on Telegram. Click here to join our channel and stay updated with the latest Biz news and updates.

Next Stories
1DUK professor in Stanford varsity’s list of top 2% scientists
2Pandemic boost: Number of licences for drug manufacturing units, medical stores soars in Gujarat
3What is La Nina and why India may witness coldest winters in January, February 2022