Thinking like a Bayesian - Julia Galef (Strata + Hadoop World 2016)

3-Second Blueprint (English)

1. Bayesian updating adjusts belief probabilities iteratively as new evidence arrives, just as spam filters weight token likelihoods.
2. Evaluate evidence diagnostically by asking how different the world would look if your current hypothesis were actually false.
3. When an observation is equally likely under both competing hypotheses, its likelihood ratio is ~1 and warrants zero confidence shift.