# Bayesian thinking *Start from the base rate* --- Evidence doesn't give you the answer. It moves the answer from wherever it was before you looked, and if you skip the starting point, the evidence misleads you. The machinery has three parts: a prior (what you believed before the new information), a likelihood (how probable this evidence would be if your hypothesis were true), and a posterior (what you should believe now). Formally: $P(H|E) = \frac{P(E|H) \times P(H)}{P(E)}$ The formula is simple. What it enforces is not: before interpreting any evidence, state what you believed beforehand and why. A 95% accurate test sounds decisive, but if the condition affects one person in a thousand, a positive result is still overwhelmingly likely to be a false alarm. The prevalence matters more than the accuracy. [[Bases]] walks this through hiring, where improving the candidate pool does more than improving the screening. Three working habits fall out of the same idea. Anchor to the base rate before the story: whatever the pitch says, start from how such things usually go, then ask what makes this one different. Ask what evidence would actually discriminate: when two explanations fit the same facts equally well, go looking for the observation that would come out differently under each, because evidence that fits everything tests nothing. And update at the rate the evidence deserves, not the rate your attachment prefers - beliefs are probabilities you revise, not positions you defend until proven wrong. The last thing a stated prior buys you is honesty. Without one, new information becomes confirmation bias with better vocabulary: the positive result confirms your thesis, the negative one was underpowered, and both feel reasonable because neither is anchored to what you believed before you looked. ---