|Decision making and 'Noise'|
I’m sure that during my professional life my decisions always were always of the highest quality and totally consistent. At least, that’s what I’d like to think! The reality was, I imagine, somewhat removed from the ideal. I don’t think that my judgements on the law were far out, but being a lawyer is also about running a business. And it is there that problems can arise far more easily.
The more people who are involved in the business, the more there is room for difference over how it should be run and variation in the performance of those employed. Lawyers vary in their ability – not so much in giving the right answer, but in actually getting the job done or being better at dealing with clients than others. When it came to the partners who were ultimately responsible for the business, we sometimes had significant differences. Should we expand and if so where? Should we instead close branch offices and concentrate on our main office? Which types of legal work should we try to develop? When we had a vacancy, whom should we hire? And of course none of these decisions had easy answers.
The latest book by Daniel Kahneman with fellow psychologists Cass Sunstein and Olivier Sibony, is called ‘Noise’. Confession first: I have not read it. But I have heard two half hour interviews with Kahneman and read some of the longer reviews, so I have a reasonable grasp of what they’re saying. Noise in this context is unwanted variability in making decisions. Their research shows that the most banal things cloud even the best expert’s judgment.
Daniel Kahneman, with his previous book ‘Thinking, Fast and Slow’ exposed the complex interactions between our rational and intuitive minds. Our intuitive minds react quickly to a situation, in effect proposing to the rational brain a possible solution for it to chew over if there is time to reflect. The research though showed that, mostly, the rational brain just accepted the intuitive brain’s suggestion and acted on it without thinking it through at all. It seems that for much of the time, we don’t really know why we do what we do.
Noise takes this view of the irrational human world further again. We have long known that decision-making is influenced by irrational and irrelevant factors. The seriousness of this was spotted in the US legal system in the 1970s. Judges were giving wildly different sentences for the same crime. Two men, for example, were convicted of passing counterfeit cheques for similar amounts; one was sentenced to 30 days in prison, the other to 15 years. The assumption was that such errors were caused by bias. Perhaps some judges were racist and gave heavier sentences to black people, or had a particular revulsion for counterfeiters.
But many more errors, it turns out, are caused by apparently unrelated random events and influences. The main culprit is what the authors call “noise”. This is a term familiar to anyone involved in electronics. The transistors, valves etc. all produce electrical noise - random electrical signals - as well as doing their job of amplifying the signals they are supposed to handle. For the authors’ purposes it means random influences on decision-making, as distinct from bias.
The authors highlight vast amounts of noise in our lives. Noise causes doctors to make different diagnoses when faced with the same symptoms. It makes almost all professional forecasters terrible at forecasting, whilst believing that they’re brilliant. Annual assessments of employees are poor to useless. Fingerprint analysis is far from the infallible system we have been told to believe in. Doctors send more patients for cancer screening in the morning than the afternoon. The weather influences judges. Better weather produces lighter sentences, as does being sentenced after lunch.
Job interviews seem to be barely worthwhile - in one experiment some interviewees were told to give random answers to questions. Not a single interviewer noticed. This is consistent with another finding – that interviewers make up their minds about their interviewees in the first 2 or 3 minutes and then spend the rest of the time selectively listening for, and so mishearing justification for that initial judgement.
University admissions are influenced by the weather, with admissions officers paying more attention to academic ability on cloudy days and more to non-academic attributes on sunny days. “Wherever there is judgment,” the authors write, “there is noise.”
Noise can also be costly in a very direct sense. The authors assessed the performance of underwriters in an insurance company. They were expected to set the same premiums for similar policies. They didn’t. Their decisions were a lottery, a noisy, random scatter, with up to 50% variation, that was costing the company hundreds of millions of dollars.
What, then, must we do? For companies the authors recommend “noise audits”, such as they carried out for the Insurers, and strategies of “decision hygiene”. Since the discovery of wide variation in sentencing decisions, sentencing guidelines have been introduced in many countries, including the UK, and appeals can be based on failure to follow the guidelines. Judges and doctors should use checklists when making decisions.
But it means also that unstructured job interviews must go, to be replaced by sessions designed to elicit relevant information from the interviewee before any members of what should be a panel of interviewers expresses an opinion. Good luck with that!
As far as business strategy meetings are concerned, there has to be an entirely different approach. At the moment, the person who speaks first tends to win the day – and of course the person who speaks first tends to be the person in charge. What is proposed instead is that each person responds to each agenda item - in writing - before the meeting and before seeing what anyone else has to suggest. All those responses should be read by everyone before the meeting so that the meeting itself becomes a discussion of all of those views. This means that the boss will have less chance of forcing through pet ideas. Ouch.
Although the idea makes us uneasy, the authors prefer the use of algorithms in principle for the making of decisions. They say that algorithms are better - less noisy - than human ones – they don’t look out of the window or have lunch - although subject to the bias of those collecting the information on which the algorithm is trained. The problem is that mistakes by an algorithm anger people more than a human error: an algorithm would have to be a far better driver than a human in order to be put in charge of a car.
But we read quite often of the success of algorithms in interpreting medical test data. They are compared to the interpretation of the data by acknowledged experts in the field and, when at least as good, are then introduced to act as a check on other consultants’ decisions. Although algorithms are not very popular, it is obviously beneficial to use them for medical diagnoses, where errors can produce tragic outcomes.
I do wonder, however, how important it is to avoid ‘noisy’ decision making in business. It presumes that there is actually a knowable best decision at the time it is made. If there is not, then the extra time and bureaucracy spent in trying to avoid randomness will be wasted. My own relatively haphazard hiring of staff seemed to work in practise, at least most of the time. In any event, there was often not much choice. Whether we would have made more money by making our business decisions only after we had first all made our suggestions in writing must be open to doubt. Sometimes, you just have to take a gamble!
Of course, I no longer have to make business decisions of any substance which affect others. Which means that I can live in what the authors call the “Valley of the Normal”. There, they acknowledge that noise in the form of human variation has a positive, creative value. I’m not sure that I like the idea that my thinking is very much influenced by random variations in my circuits, but I suppose that it’s something I shall have to live with.
18 May 2021