Statistics and the answer to everything
Published on Tuesday, 07. September 2021Forty-two. The answer to the ultimate question of life, the universe, and everything. I find it much more interesting that it's also a sphenic number. And actually, this is my 43. post. Starting to count at zero can be confusing sometimes.
The idea of an answer to everything is somewhat ridiculous. I'm not talking about a theory of everything here. A theory of everything is a framework that fully explains all physical aspects of our universe. And while such a theory wasn't yet found, it's possible (and I think likely) that it exists. But it isn't a truth-finding mechanism to find the answer to everything. While it might have some surprises like faster than light travel in stock, it won't help you to decide whether you'll win the next round of poker or whether you should marry someone. An answer you can use for everything is either overly generic that it isn't helpful or is simply wrong at least some of the time. It's an excuse to stop thinking.
Let's assume you want to figure out if a media outlet is saying the truth. One possible answer is that it's always telling the truth. This is unlikely to be correct. An equally wrong answer is to say that the outlet is always wrong. If you are thinking this, your opinion is equally informed by the outlet, it's only inverted. Neither of these answers are a valid truth-finding mechanism. The hard answers that are more likely to be true include "I think they get a lot of things right, but are wrong on this issue" or "I think they are right here, even if I don't like the way they came to this conclusion".
One pervasive non-answer of our time is hidden behind the mask of empiricism. On the surface "I need more data to think about this" sounds like a reasonable answer. But data isn't the answer to everything. It can be incomplete or interpreted differently. The same data can support different, mutually exclusive theories. So to verify a theory, you need to make predictions with it and check it against data. But you don't need more data to come up with a theory. A better answer is to think about what theories are possible with the current data you have, and to write down conditions under which they are no longer valid. Once you see evidence that your theory no longer works, it's time to change your mind.