Bayesian Updating Interview Drills
Bayesian updating interview drills for priors, likelihoods, posteriors, count tables, signal reliability, and base-rate mistakes.
Candidates who know Bayes rule but need repetition across signal and base-rate cases.
Name the prior
Bayesian updating starts with a prior probability before the new signal. State it clearly so the update has a baseline.
Add the likelihood
The likelihood says how probable the observed signal is under each possible state. Interview mistakes often come from swapping P(signal given state) with P(state given signal).
Use count tables
When percentages feel slippery, convert to an imaginary population. Count how many cases are in each state and how many produce the signal.
Concrete drill
If 1 percent of items are special and a test catches 90 percent of special items while falsely flagging 9 percent of ordinary items, compare true flags with false flags before giving the posterior.
Signal reliability
A reliable signal can still produce a modest posterior when the base rate is small. The base rate and likelihood must both be included.
Common mistakes
Candidates often update only with the test accuracy and ignore how common the underlying state was before the signal.
Practice the pattern
Use the LeetQuidity curriculum and calibration to turn this topic into a focused practice plan.