Kahneman for Thugs: some bullet points

So I did a business-class review of Daniel Kahneman's new book over at the Fistful. Of course, AFOE is a very different blog to this one, being all liberal euro-technocratic and whatnot. Therefore I thought I'd write a different review. I therefore give you, in this week of Rupert, Kahneman for Thugs - which is at least appropriate given how much work he put into improving the Israeli army's officer selection process. In this post, I'm going to be deliberately political and action-oriented, and I'm going to express myself in handy bullet points.

Availability

The first point is that rebuttal is futile and you can ignore nothing. The so-called psychological anchoring effect means that our intuitive judgments of anything are influenced by whatever information is around at the time. That includes information we know to be false and information that is completely irrelevant. Ask people to guess the weight of a pig, and they will guess higher if you mention a big number beforehand.

Also, intuitive judgments of truth are strongly influenced by cognitive fluency. Things that are easy to remember are true. Things that fit into a coherent story are true. Things that avoid conflict are true. This has some truly weird consequences - if you want people to learn something from reading a text, choosing a font that is hard to read actually increases how much they retain. However, if you want them to believe it and act on it, you want to be nice and sans-serif.

So, you can't ignore anything, repetition works, and getting in first works. Smear politics is effective. On the other hand, the only way you can prevent Andrew Gilligan from influencing your opinion of Ken Livingstone is to just stop reading him. Filtering (on your part) or censorship (on their part) is effective.  

Substitution

Thinking is work, and as a result, people unconsciously try to answer the easiest question that seems to fit. It's therefore important to a) set up problems so they answer the question you want them to, and b) set up your own intuitions to work with your own interests. If you train yourself adequately, all you will hear from Simon Jenkins is "blah blah blah fishcakes" as Boris Johnson memorably said. Politics as a system of grudges is an effective way to operationalise the points above.

Surprise

Repetition, fluency, and availability dominate what we believe. But this is only half the point. What if you need to convince? The answer is surprise. Uncertainty is information, and learning is the process by which that information is trained into the near-automatic activity of System One. To surprise is to convince. This doesn't contradict the point about cognitive fluency - the point is to create a coherent story that contains a surprise and therefore a change in opinion.  

Depletion

Thinking is work. It requires effort, unlike intuition, which is associated with no sense of subjective effort and no change in psychophysical metrics. Therefore, as you get tired (or hungry, or drunk, or ill) you rely more on intuition. Judges are more likely to give you parole early in the morning (after breakfast) or early in the afternoon (after lunch). Therefore, it is effective to target the depleted and deplete the targeted. What did you think all the anti-design graphic noise on the front of the Sun, or Fox News's horrible screen graphics, are for?

Never quote base rates

People find it hard enough to make judgments that involve statistics, and tend to neglect denominators. You can help this process, by never providing base-rate information. Certainly, if you want to mention how many terrorist attacks, crimes, or whatever happened this year, don't say how many there were last year unless that number was unusually low. And never, ever quote more than two data points.  

Never question premises or permit them to be questioned

Many cognitive biases seem to disappear if the decision involved is taken in a wider context. Obviously, if you can control the context (see under Availability) that's all to the good, but this is rarely possible in a pluralistic society. Very often, though, people take decisions without using any external reference, and over-focus on the exact terms of the question (known in the tech industry as bikeshedding), which of course means over-focusing on the easier question they substitute in. This inside view effect is a powerful source of error. Cherish it.

On the other hand, it's almost a cliché that mediocre candidates answer the exam paper while brilliant ones question it. Disputing the terms of the question is an effective defensive tactic.  

Always isolate questions

Many cognitive errors that appear when choices are isolated disappear in so-called joint evaluation. If the choices are isolated, people often make decisions which are mutually inconsistent. Very often, if you put the questions together, they succeed in integrating the information involved into a common picture. Therefore, it is necessary to isolate questions and prevent joint evaluation. Whether Saddam Hussein is a bastard or not must be isolated from the question of whether there are ways other than war to limit the consequences of this bastardy and from the question of what the costs of the war might be.

All clear so far?