More tales from the joy and delight that is teaching principles of economics.
I explained what Goodhart’s Law is in class the other day, and if you’re wondering what it is, here you go:
Vishesh emailed to ask a very good question (note that I have paraphrased the question to make it easier to understand, this is not verbatim):
Say you want to be more regular at your gym. In order to build this habit, you decide that you will treat yourself to an ice cream after you go to the gym.
At the outset, you want to go to the gym because you want the ice-cream, not because you want to go to the gym.
But eventually, you will start to want to go to the gym for its own sake, rather than for the reward of the ice-cream. The ice-cream won’t matter, managing to deadlift a 100 kilograms will matter more.
So is Goodhart’s Law not applicable here?
My answer to the question went something like this:
Well, if going to the gym has become important, you no longer need to measure it, and so yes, Goodhart’s law is no longer applicable. Note that it hasn’t failed – it is not applicable. Why is it no applicable? Because you aren’t measuring gym attendance, or targeting it. It “just happens”, now for its own sake.
That’s one way to explain it. This is the better way: