This one adds to my emerging thinking that one of the principle difficulties we face in conducting unbiased research is that we're swimming against a biological imperative. Humans are just not wired to operate in an unbiased way. We are, in fact, wired to fool ourselves and presumably at a deeply instinctive level.
With a lot of evidence that erroneous beliefs aren’t easily overturned, and when they’re tinged with emotion, forget about it. Explaining the science and helping people understand it are only the first steps. If you want someone to accept information that contradicts what they already know, you have to find a story they can buy into. That requires bridging the narrative they’ve already constructed to a new one that is both true and allows them to remain the kind of person they believe themselves to be.
The comment above is framed as if a scientist is trying to persuade a lay skeptic. But as scientists, these are conversations we need to also have with each other and particularly with ourselves. We have to find a way to cut through what seems to be a deeply rooted psychology that functions to convince ourselves that the decisions we make are right, even when we are dead wrong.
But somewhere along the lines over the last couple of decades we stopped using statistical procedures for the reason they were designed, which is simply to keep our thinking and the conclusions we draw from our experiments as unbiased as possible.
The practice switched from conducting unbiased statistical tests of hypotheses to running the calculations as affirmations of the significance of our ideas. That's a major philosophic distinction. Statistics morphed into a cloak that we wrapped around our brilliant self-deceptions. The situation we grew into is not unlike the Emperor's new clothes. We fool only ourselves when our intention is to test how right we are.
Yes, we've made tremendous progress and fantastic discoveries over the last 20 years. But the landscape is riddled with irreproducible results.
Bitter irony, the reason I was poking around on the 538 site and stumbled into this was to analyze the wreckage of my NCAA tournament picks, which all seemed so brilliant prior to the event. I'm in an auction-based game, where half the pot goes to the most points. You earn points equivalent to your team's seed each time they win. I did a probability-driven play selecting a stable of 8-9-10 seeds. This worked brilliantly in round 1 (50 pts!), but I emerge from round 2 with only a single pony in the race. I figured I needed at least two 2nd round upsets for a shot at the money.
Go Syracuse! You are all I have left.