A formal framework for choosing among actions by combining prior probabilities, observed evidence, and the value or cost of each possible outcome. Rather than treating beliefs as fixed, a Bayesian decision-maker updates them as new data arrives and selects the action that maximizes expected value under the updated distribution.
This framework excels in environments with structured uncertainty — pricing experiments, clinical trials, capital allocation under risk. Its main cost is requiring the decision-maker to make implicit assumptions explicit: priors, likelihoods, and utility functions all have to be written down.