Definition
The expected utility of an action is the probability-weighted average of the values of its possible outcomes. For each outcome you multiply how good it is by how likely it is, then add the products together.
In decision theory the expected utility serves as the criterion of rational choice: of the actions available to you, the rational one is whichever has the greatest expected utility. The rule combines two ideas — desirability and probability — into a single comparable figure.
Why it matters
How it works
Suppose an action might lead to outcome A worth 10 units with probability 0.3, or outcome B worth 2 units with probability 0.7. Its expected utility is 0.3 times 10 plus 0.7 times 2, which is 4.4. Compare that figure with the expected utility of every rival action and choose the largest.
Priest distinguishes expected utility from raw expected monetary value. Because the value money carries is not linear — a second million matters less than the first — the same cash payoff can have very different utility for different agents. This is why utility, not money, is the quantity the rule averages over. The deeper worry is that, in cases like the Newcomb problem, maximising expected utility recommends an action that a competing dominance argument rejects.