new york times

A Test of Confirmation Bias

The New York Times posted a quick puzzle test which you can try here. If you enjoy this sort of thing, it's a fairly standard number puzzle you may have seen before. But the results are extremely counterintuitive. Go try the test and see how you do before reading any further...

I'll wait... Promise.

I've known the answer since university (Math professors throw these things out all the time) and it highlights that the way most of us go about trying to solve problems the wrong way.

Everyone usually starts out right trying to generate some possible solutions - guesses at what the right answer might be. But then when it comes to choosing the best one, things get messy. The common belief is that we search for evidence which confirms our theory. It doesn't work, since one piece of evidence can be consistent with many different possible solutions, the fact that the piece of evidence agrees with any one solution, doesn't help picking one solution over another. Counter-intuitively, piling on more evidence that agrees with your hypothesis doesn't help distinguish one solution over another; it doesn't move you forward.

The correct path is the opposite direction. You go out in search of evidence which would go against your hypothesis. And if you try hard to find it and come up empty, then you can be confident. When it comes to deciding, one piece of inconsistent evidence is more valuable then a thousand pieces of consistent evidence.