Friday, February 29, 2008

Parenting superpowers

Sometimes the mere application of statistical knowledge can give you abilities that seem super-normal to children. Let me take one example. I send my son upstairs to get dressed. Five minutes later I shout out "stop doing that and get dressed." My son comes downstairs dressed and asks, "how did you know I wasn't getting dressed?" I respond: "I see all, don't forget it."

In fact, I only see probabilities and let me tell you that there is no safer bet that my son has "got lost" before he has managed to get dressed. So an occasional shout out based on that bet and no observed evidence is a pretty safe one.

But this form of statistical superpowers doesn't always work: it may be a safe bet but it is still a bet. Let me take another example attempting to use a far more sophisticated super-power: statistics + game theory.

We came home the other night after going to our daughter's parent-teacher interview. Immediately, she bounded up to me and asked us how it was. Hmm, how unusual? I thought. Why is she so interested?

"So how did it go?"

"Well, you know?"

"I know what?"

"He told us about the ... incident."

"The incident?"

"Yes, would you like to tell me your side of the story?"

Sadly (for me at least), it transpired that there was no 'incident' (my supposition of poor behaviour or trouble) and that she was just interested in what the teacher had to say. There was no information to be extracted so freely. I maintain it was a good bet but the sad thing is that you can't play that card too often: statistics persists but a game theoretic advantage evaporates. I'll have to leave that one for a few more years.