Book Review: Thinking, Fast and Slow by Daniel Kahneman
29 January 2014, by Chris Harris
This is a book about how we think: how we assess probabilities, form opinions and make decisions. Its central theme is that there are a few specific ways in which we are a lot worse at this than we imagine, and there are some techniques we can use to counteract this. It’s an excellent and entertaining book which everyone should read.
On the face of it, it isn’t that relevant to software development, but since I started it a month or so ago I’ve already found a number of applications during the working day – particularly to project management. I’ll start with a typical experiment to give you a flavour of the book, and then give three findings that I think are applicable to our field.
Linda the bank teller
Linda is 31 years old, single, outspoken, and very bright. She majored in philosophy. As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in anti-nuclear demonstrations.
Which is more probable?
- Linda is a bank teller.
- Linda is a bank teller and is active in the feminist movement.
In the initial experiment, 90% of those asked chose option 2. And even now option 2 feels more palatable to me; however, hopefully you can see on reflection that option 1 must be more probable, as it is a superset of option 2. This specific fallacy is called the conjunction fallacy and it has plenty of more practical, albeit less striking, consequences. Kahneman argues that it is caused by the representativeness heuristic, whereby we judge the probability of something being true by forming a picture of it in our minds and comparing it to reality. This works well a lot of the time but, as we can see, it sometimes gives the wrong answer.
Kahneman recommends exploiting the representativeness heuristic (and other related phenomena) by running “pre-mortems” when trying to predict the success or otherwise of a project. We already do something similar at Softwire (“risk brainstorms”, or “risk storms” for short), but here, instead of asking “what are the risks on this project?”, you say “Cast your minds forward to the expected end of this project. Imagine it has gone badly wrong – one or more of the chief objectives has not been met. How did that happen?”. Suddenly it’s very easy to picture the scene and work your way backwards to see all the things that could have gone wrong on the project (see also Outside View below).
I’ve tried this out a couple of times on my projects since, and the results were very promising. I’ll probably write a separate blog post about them at some point.
If you put someone in a position of discomfort (say, put their hand in very hot water) for half an hour, and then you do the same again, but add on an extra 10 minutes of slightly milder discomfort (say, slightly less hot water), and then ask them which of the two experiences they’d like to repeat, they will generally go for the latter option. That is, they’ll take the option that involves more discomfort overall, so long as their last experience was less painful.
This is not because we’re masochistic, it’s because we’re not very good at remembering how much we liked or disliked something. For one thing, the amount of time spend experiencing discomfort and pain is almost entirely disregarded. For another, we place more weight on the most recent aspect of the experience.
In another experiment, Kahneman asked students to fill out a questionnaire, asking them among other things “how happy are you?” and “How many dates have you gone on in the last 3 months?”. There was a very high correlation between the two answers – but only if the dates question preceded the happiness question. When the happiness question was asked first, there was little if any correlation between that and the number of dates they’d been on. It seemed like the students who were primed to think about their love lives used that as a “substitute question” for the harder happiness question.
His broader theory is that everyone, when asked a hard question like “how happy are you”, substitutes it for an easier one and answers that instead. I have long been of the opinion that polling people with explicit questions such as “how was your day today”, “how happy are you on your current project”, etc. is fairly useless, for these kinds of reasons.
One conclusion to draw is that our memory of a project or task will be influenced by a) how it ended, and b) how much good it did us in the long run, and so we should be working towards improving these (through hitting our goals and learning new skills) rather than trying to improve a daily happiness score. Maybe we can then give ourselves easier and more useful question to answer, such as “how much did I learn today?”
My favourite anecdote in the book is a cautionary tale about our inherent optimism. Kahneman and some fellow experts had persuaded the Israeli government to include a course on decision-making in schools, and had been charged with writing the textbook. A few months in, they decided to try and estimate how long it would take. Their estimates ranged from about 18 months to 2 1/2 years. Then Kahneman asked the printer, who had been involved in many similar projects before (and who had estimated about the same as everyone else), how long the average such project took to finish. He thought for a while and then conceded that, among those that had actually finished, the range was probably about 7 – 10 years. When pressed further he admitted that from their position, only about 40% of the projects ever finished at all.
It’s interesting that the printer had had to take the “outside view” in order to provide a decent estimate. If he hadn’t been asked the question, he’d never have made use of the valuable experience he already had. But even more disconcertingly, this outside view was simply brushed under the carpet by the group (all leading experts in decision-making, remember) – it was an unthinkable outcome and so they just didn’t think about it. Based on these odds, most of the team would have bailed out immediately, but they pressed on, and finished 8 long, dispiriting years later. The textbook was never used…