Base120 · Recursion
Bayesian Updating in Practice
Continuously revise beliefs as new evidence arrives, weighting by reliability
When to use
"Use when beliefs should be revised as evidence arrives. Start with a prior, update with new evidence weighted by its reliability, arrive at a posterior. The posterior becomes the prior for the next update."
Example
"Prior — 30% chance the feature will succeed. New evidence — early user testing shows 80% task completion (strong positive signal). Update — posterior rises to 65%. Next evidence — beta users abandon after 1 week (strong negative signal). Update — posterior falls to 25%."
Common misuse
"Updating beliefs without weighting evidence by reliability. A single anecdote and a randomized controlled trial are both evidence, but they should not update beliefs equally. Weight by sample size, bias, and methodology."