The Drunkard’s Walk


I’ve just started a book called The Drunkard’s Walk: How Randomness Rules Our Lives. It’s quite interesting. Very early on, there’s a discussion about whether yelling at people when they do badly or praising them when they do well works best. The people who do the yelling or praising are under the impression that yelling works better. After all, if someone does something they should know how to do but does it exceptionally badly and then gets yelled at for it, the next time they do it, they are almost always better at it. On the other hand, if someone does something they should know how to do but does it exceptionally well and then gets praised for it, the next time they do it, they are almost always worse at it.

Having worked with training dogs much of my life, I’m very clear that yelling when you are trying to train does not actually help, so it surprised me to see this, at least until I read on a bit. People vary around a mean, generally. On average, they get better with practice, but they will have good days and bad days as the skill level slowly increases. The average will increase very slowly, though, so very good days or very bad days are unusual.

Why does yelling seem to work and praise seem not to work? Because people will generally go back to their normal skill level, whether it’s above or below the notable performance. If they have a really bad day, whether or not someone yells at them, the next day is likely to be better. If they have a really good day, the same applies – praise them or not, the next day will probably be worse.

One of the ideas in the book is that human brains aren’t set up to handle random events. We try to see patterns even where there aren’t any. We’re back to “correlation does not indicate causation”. We may see something that seems to indicate that yelling at people improves their performance, but it doesn’t. The person being yelled at is simply coming back to what is normal for them.

If someone has a daughter, what is the likelihood that the second child will be a daughter? I would have answered 50%, but it isn’t. It’s 33%. The possible arrangement of children if one is female is girl-girl, girl-boy, and boy-girl. Once you know that one child is female, it changes the odds.

We aren’t good at randomness, and intuition doesn’t help us when we are looking at random events. It can hurt us, even, by giving us information that, if we could lay out problems mathematically, doesn’t make sense.

I’m enjoying the book. It’s making me look at the world in a different way, which is always interesting, and in some ways it is calming. I worry about things I can’t control, and this is helping me remember that I really don’t have any way of predicting things. It’s a little scary to think of the world as a random place, but I know it is; I jut haven’t wanted to face up to it.

The concepts aren’t necessarily easy at times, since they are counterintuitive, but the book is written so it is easy and fun to read.