Wednesday, December 15, 2010

How do you know that?

Imagine for a moment a combination lock with 100, 1-digit numbers. There are 10^100 different possible combinations. You enter a combination at random. Is it right or wrong?

It's probably wrong. In fact, the chance that it is the correct one are vanishingly small. There are so many more ways to be wrong than to be right.

Now let's try it again. Same combination lock, only this time, you have the instruction manual which says the default combination is all zeros. You know this lock is brand new and has not been changed. You enter the all zero combination. Is it right or wrong?

It's probably right.

What changed? You had some evidence that showed that one of the 10^100 combinations was much more likely than all the others. It's still not certain (factories do make mistakes).

Now imagine that in both cases you entered a combination, proudly announcing it to an observer before opening the lock. The observer asks, "How do you know that?"

In scenario 1, you don't have a good answer. There's no reason to suspect that you have the right answer. You might make up a reason. You might say that you have the manual when in fact you do not. But if they follow this up by asking to see the manual, you can't provide it. More importantly, even if it turns out that you have chosen the right combination and the lock opens, you still got there with wrong methods that will most likely not work next time.

In scenario 2, however, you have a fine answer. You can show the observer the manual and explain your chain of reasoning. If you are in the same situation later, you can follow a similar process. And even if it turns out that the lock does not open, you used correct methods that will most likely work next time.

When making decisions, always follow the evidence. Because, like this thought experiment, we live in a world where there are far more wrong answers than right. And no amount of lies or faith will turn a wrong answer right.