Human decision-making is very unstable, with people quite often giving different answers to the same question
One of the delights, and frustrations, of dealing with people whether in business or in everyday life is that you can never tell what they going to do next.
Think how boring life would be otherwise! Imagine tennis, for example, with all the variability in human nature magically removed.
The succession of shots in each rally could be forecast in advance; each serve would be the same - perhaps relentlessly out wide, or always down the centre, and hit with the very same speed and spin; and every serve would be ‘in’ of course. Needless to say, the outcome of the match itself would be predictable and very dull.
The fundamental unpredictability of human nature plays an equally crucial role in just about every aspect of life, from the wild uncertainties of entrepreneurship and the gyrations of the stock market, to the impossibility of guessing with any certainty future fashions, fads, trends, political victories, armed conflicts, GDP growth or rates of inflation.
Where does all this variability come from? One answer, of course, is that human behaviour is usually the outcome of a complex interaction of many factors (in individual brains, organisations, or whole economies).
Small variations in any of these can lead to unexpected and large changes in the outcome.
A second reason is that sometimes unpredictability is crucial to success. Tennis players want predictably to hit their shots ‘in’ - but they want to make their choice of shot as unpredictable as possible. In the same way, unpredictability can be crucial in outwitting the competition in business.
But looking closely at human psychology, the sheer randomness of behaviour seems so widespread that it seems it must have an even more fundamental origin.
Suppose, to pick a standard and much-studied type of lab task, people are asked whether they’d like to play a lottery (with a known chance of winning a large-ish sum of money) or whether they’d prefer to ‘play it safe’ (taking a smaller amount of money for sure).
Psychologists and economists have used tasks like this, with real money, to work out how much risk people are willing to take, how they balance gains and losses, and much more.
But a really striking finding, often ignored, is that people’s choices are incredibly unstable. If asked the very same questions twice in the course of an hour’s session in a lab (mixed together so that they are unlikely to notice), people quite often give different answers - indeed often around one quarter of the time.
It is the same with surveys: market researchers often ask people the same question several times in different ways and take the average.
Here’s a particularly extreme example of variability gone ‘wrong’. Take a biased coin, which comes up heads two-thirds of the time and tails for a third of tosses. The task is to successively to predict the next coin flip.
Typically, people mostly guess heads (in fact, about two-thirds of the time); but sometimes guess tails (about a third of the time).
A stream of guesses might be: HHTHHHTTHHHTTHH… (notice people tend to favour irregular ‘random-looking’ orders - but biased towards heads of course).
But this is completely the wrong thing to do. Every choice is really the same choice: there is always a 66 per cent chance of being right if you guess heads and a 33 per cent chance of being right if you guess tails (you’ll win five times out of nine if you do this).
So you should always guess heads and you’ll win two thirds of the time, which is six times out of nine. Yet even in this incredibly simple task, almost nobody does the ’right thing’.
It just seems too weird just to guess HHHHHHHHHHHHHH… indefinitely. The human urge to throw in some variability seems to overwhelm us.
How do people make decisions?
In short, human behaviour is noisy, even when it shouldn’t be. Indeed, high levels of noisiness seem to be built into the way our minds work. So, what is going on?
Together with Adam Sanborn, of the University of Warwick’s Psychology department, and a brilliant team of post-docs and PhD students, I’ve been exploring a new explanation of human unpredictability, which assumes that noisiness is not a ‘bug’ that needs to be eliminated, but is essential to how the mind works.
The essential insight is that the brain needs to deal with an uncertain world, but can’t possibly do the complicated mathematical calculations demanded by probability theory.
But AI and statistical research confronts this problem daily: almost all real-world problems are too complex to calculate out precisely, so they’ve devised 'tricks', justified by elegant mathematics (so-called Markov chain Monte Carlo, or MCMC), to quickly get approximate answers. Perhaps these tricks are also used by the brain.
To get an intuition for how this might work, suppose you hear that an ‘animal’ has escaped from a nearby zoo. Should you be worried? Well, figuring out the probability that the animal is a tiger/lion/giraffe/koala etc is way too difficult (there are so many animals!), let alone how dangerous an encounter might be.
But the brain may be able to give us a rough estimate by randomly sampling a few animals (for example a lion, tiger, leopard…), roughly in proportion to their probabilities, and making a decision using just these examples.
The theory of MCMC shows how, with the right random sampling method and enough samples, this strategy can work very well (the samples will eventually be a good representation of the entire probability distribution of animals).
But our brains mostly don’t have time for that. We hastily draw a few samples, and act accordingly. Of course, which random samples we draw will matter. Sampling lion, tiger, leopard may lead us to hurry fearfully indoors; sampling mongoose, tapir, and platypus probably won’t.
So this is where unpredictability comes from. We think about uncertainty by filling in a few specific scenarios and reasoning as if one of those will come true. Different random samples will lead to different choices.
The same point applies everywhere. Thinking about current economic woes, we might draw scenarios based on past recessions - some indicating a lasting malaise, others rapid recovery.
Wondering about the implications of a new infectious disease, we sample memories of similar outbreaks. But the samples will be different between people and between when we think about it. A business proposition that seemed a sure-fire winner one night might, when we recall a few failed start-ups, feel doomed to fail the next morning.
The ‘brain as MCMC sampler’ viewpoint can be turned into a specific computational model of the mind, with a wide range of predictions, about how we retrieve information from memory, judge probabilities and the fine-details of the timing of behaviour of all kinds.
We’ve been building models, testing, and refining them for nearly a decade now. Intriguingly, some of these patterns are strangely reminiscent of statistical patterns in the stock market (possibly suggesting that the psychology of individual traders might be ‘leaking through’ to overall market behaviour). It turns out, too, that this viewpoint has independently been proposed by neuroscientists, trying to understand noise in brain activity.
But as we’ve seen, the most fundamental consequence of our approach is that it explains the otherwise puzzling unpredictability of human behaviour.
Noise isn’t a nuisance, to be eliminated. Instead, randomness lies at the very core of how our brains deal so successfully with a complex and uncertain world.
Nick Chater is Professor of Behavioural Science and author of the award-winning The Mind is Flat and co-author of The Language Game. He teaches Judgement and Decision Making on the MSc Finance, MSc Accounting & Finance and MSc Business & Finance. He also lecture on Behavioural Sciences for Managers on the Distance Learning MBA.
Follow Nick Chater on Twitter at @NickJChater.
For more articles on Decision-Making and Analytics sign up to Core Insights here.