Thinking, Fast and Slow, By Daniel Kahneman

Behavioural economics: The varying speed of judgement may depend on lunch (Rex Features)

How fine it must be to have Daniel Kahneman's confidence in the effect of one's words! "Your face twisted slightly in an expression of disgust, and you may have pushed this book imperceptibly farther away," he reports, having presented the word "vomit" to his readers. "Your heart rate increased, the hair on your arms rose a little, and your sweat glands were activated." He seems to know you better than you know yourself.

And indeed he does, to an extent recognised in 2002 by the award of the Nobel Prize in economics. Kahneman and his intellectual partner Amos Tversky, who died in 1996, humanised economics by demonstrating systematic biases in the way people assess prospects and make decisions. They offered evidence about mental processes in place of the economists' traditional assumption that humans make their choices by thinking rationally. Now their view of the mind has become conventional, and that passionless accountant homo economicus survives only as a straw man who illustrates the superior understanding of psychologically informed "behavioural" economics.

Since Tversky and Kahneman published their first landmark paper (an appendix to this book) in 1974, the map of reason's hinterland has become more detailed. Kahneman resolves thinking into two systems. One is intuitive and emotionally inflected, which makes it a fast and generally effective "machine for jumping to conclusions". The second is slow and rational, offering a means to check the responses made by the first system, and to understand the world more rigorously than intuition allows. However, this second system is not only slow but lazy. It's inclined to go along with whatever the first one suggests.

To promote slow thinking, Kahneman urges his readers to work through the example problems with which he sketches the landscape of thought. It feels like work to the body as well as to the mind. "Your muscles tensed up, your blood pressure rose," he tells us. Alternatively, your cognitive lethargy protested at the challenge, and your eyes glazed over. Jumping to conclusions takes a lot less effort.

Two fundamental features of the fast system are that it is averse to losses and it can't do statistics. To intuition, "losses loom larger than gains". Bad things in general appear to impress us more than good ones. The psychologist John Gottman reported that for a relationship to remain stable, there have to be at least five times as many good interactions within it as bad ones. Even in passion there is accountancy, and the negative scores weigh heavier than the positives.

Mental book-keeping is an essential skill in an intensely social species full of individuals constantly doing favours for and taking advantage of each other. This is very different from statistical thinking, though. We are always looking for causes – particularly human ones – and we are only too ready to conclude that one event has caused another when we see an association between the two.

Nor do we readily grasp what randomness looks like. During the Second World War, the German bombs missed some parts of south London, causing people to suspect that enemy spies were hidden in those quarters. Statistical analysis, however, showed that what looked like a pattern was probably the result of chance.

In situations where people are striving to affect events, other possible influences fade into the background. During Kahneman's intellectually formative years as a military psychologist in Israel, he encountered flying instructors who insisted that praising good performance was counter-productive. They had observed that cadets praised for executing a manoeuvre well would generally do it worse the next time they tried it, whereas cadets berated for making a mess of it would do better next time. They failed to appreciate that by the nature of variation, a notably good or bad attempt would usually be followed by one closer to the average: worse if the preceding one had been good, better if it had been poor.

A more recent observation from Israel, on parole judgments, casts a more unsettling light upon the reasoning faculties of individuals in positions of authority. Prisoners' chances of parole reached about 65 per cent after the judges had eaten a meal, but dwindled towards zero by the time the next meal was due. The explanation was not that hunger makes judges vindictive, but that reasoning demands energy. Overall, only about 35 per cent of requests were granted, so refusal was the position to which the judges defaulted as their batteries ran down.

Worrying as it is for our ideals of justice, this finding illustrates nicely why fast thinking comes first and slow cogitation comes second. The judges became tired and hungry. Much of humankind has spent much of its existence tired or hungry or both, and under such conditions people are particularly vulnerable. Fast thinking enables people to get by when they can't afford much in the way of reasoning. Making a coherent story from the information to hand may have to suffice.

Fast thinking is not just a means to cope with scarcity or crisis, though. Kahneman lauds its general accuracy, efficiency and flexibility: if it can't answer one question, it will find an easier one and answer that. But he admits that the automatic thinking system is "not readily educable". The times when you need fast thinking the most are precisely the times when you can't stop to check whether you are falling into cognitive error.

If you know what to look out for, though, you may have a much clearer view of other people's mistakes. Kahneman offers his readers a vocabulary to describe the biases of thought, in a text that surveys both his own work and the rich field it has seeded. It sometimes requires more work than it should: at several points I had to turn to the internet to find clearer accounts, including by Kahneman himself. But that is because much more than a single lifetime's slow thought has gone into it.

Marek Kohn's latest book is 'Turned Out Nice' (Faber & Faber)