The Value of Information

Before you pay for one more test, study, survey, or expert opinion, there’s a single question that cuts through almost every “should we gather more data?” argument — and almost nobody asks it: could the answer actually change what I do? If you’d make the same move no matter what the test says, the test is worth exactly zero, however accurate, cheap, or fascinating it is. All the value of information comes from the mistakes it lets you avoid.

This is the value-of-information model, straight from decision analysis (Ron Howard, 1966). It gives you three sharp tools: the would-it-change-my-action test for spotting worthless data instantly; EVPI, the most a perfect answer could ever be worth — the ceiling; and EVSI, what a real, noisy test is worth once you fold in its errors and subtract its price. It’s the model that tells you when to look before you leap — and when looking is just expensive procrastination.

This is an expert-tier course. It assumes you can already revise a belief from evidence (Bayesian updating), weigh outcomes by their odds (thinking in probabilities), and see why a convex payoff rewards keeping your options open (asymmetry and optionality). We’ll put all three to work turning “gather more data” from a reflex into an expected-value decision you can actually compute.