Value of information
Also known as: Expected value of information
How much it is worth to learn something before you decide.
What it means
The value of information is the amount by which acquiring additional information improves the expected outcome of a decision, and thus the maximum one should rationally pay to obtain it before choosing. In decision analysis it is computed by comparing the expected value of the best decision made with the information against the best decision made without it; the difference is the information's worth. A key implication is that information is valuable only insofar as it could change what you would do — perfect information about a question whose answer would not alter your choice is worth nothing, no matter how curious you are. The expected value of perfect information sets an upper bound, while the expected value of sample (imperfect) information accounts for tests that are noisy or partial, weighing the expected decision improvement against the cost and delay of gathering it. This framework formalizes when to run a study, order a diagnostic test, or seek expert advice versus act now. It matters because it disciplines the intuition that more information is always better, showing instead that information has value only when it is both decision-relevant and worth its cost.
Examples
A doctor declines an expensive test because, whatever its result, the treatment plan would stay the same — the test's value of information is zero despite its diagnostic detail.
A team spends six weeks surveying the market for a product leadership has already committed to building. Whatever the survey said, the plan was not moving, so it bought only delay.
A buyer pays handsomely for a structural survey on a warehouse, and it is money well spent: if the survey finds subsidence the purchase is off, so the answer can genuinely flip the decision.
First described in Decision analysis; Howard (1966); foundations in Bayesian statistics.