Explain any topic with three levels of analogy

*(This sample picks up after the two check-in questions, where the learner said they are comfortable with basic probability but have never used Bayes' theorem in practice.)* **For a 10-year-old — the lost puppy** You think your puppy is probably asleep under the bed, because that's where she usually is. Then you hear a crunching noise from the kitchen. You don't throw away everything you knew about her — you just become a lot less sure about the bed and a lot more sure about the kitchen. *How it maps:* your first hunch is the prior, the crunching noise is the evidence, and your updated hunch is the posterior. *Where it breaks:* real updating is arithmetic on numbers, not a feeling, and you can end up only slightly more confident rather than switching your mind entirely. **For a high-school student — the medical test** A test for a rare disease is 99% accurate, and yours comes back positive. It feels like a 99% chance you have it. But if only 1 in 10,000 people have the disease, then in a group of 10,000 you get about 1 true positive and about 100 false positives — so a positive result means roughly a 1% chance, not 99%. *How it maps:* the rarity is the prior, the test result is the likelihood, and the surprising 1% is the posterior. *Where it breaks:* it makes Bayes look like a trick for medical tests, when it is really the general rule for revising any belief. **For a college-level reader — the beam of belief** Picture your belief as a fixed amount of light spread across every hypothesis. Evidence doesn't add light; it re-aims the beam. Each hypothesis keeps a share proportional to how well it predicted what you actually saw — so a hypothesis that expected this exact observation gains at the expense of one that expected something else. *How it maps:* the fixed total is the normalising constant, the redistribution is multiplication by the likelihood. *Where it breaks:* it suggests a fixed set of hypotheses, whereas in practice the true one is often not in your list at all. **In plain terms:** Bayesian inference is a rule for changing your mind by the right amount. You start with how likely each explanation seemed, weight each one by how well it predicted the evidence you actually observed, and rescale so the possibilities still add to one.