Expected Utility¶

Origin domain: Economics & Finance
Also from: Information Theory, Security Intelligence, Biology & Ecology, Engineering & Design
Aliases: Expected Value of Outcomes, Probability Weighted Payoff, Von Neumann Morgenstern Utility

Core Idea¶

Expected utility is the structural pattern of valuing an uncertain prospect by weighting the value (utility) of each possible outcome by its probability and summing — collapsing a distribution of futures into a single comparable number that ranks choices under risk. The defining commitment is probability-weighted aggregation of a value function over outcomes, where the value function is generally nonlinear (concave for risk aversion), so that the worth of a gamble is neither its best case nor its average payoff but the expectation of utility, not of money.

How would you explain it like I'm…

Worth of a Gamble

Imagine a mystery bag of candy. One bag almost always has a small candy. Another bag rarely has a giant candy. To pick which bag is better, you don't just look at the biggest candy, you also think about how often you'd actually get one. You sort of blend the size of the candy with how likely it is, and that gives you a fair feel for which bag is the better deal.

Average Value of Chances

When something is uncertain, smart deciders combine two ideas: how good or bad each possible result is, and how likely each result is. They multiply each result's value by its chance, then add those numbers together. That single number lets you compare risky choices the same way you compare prices. It also explains why people don't always chase the biggest prize: a small chance of a huge reward can be worth less than a sure thing, because rare wins don't add up to much on average.

Probability-Weighted Value

Expected utility is a rule for ranking risky choices. For each option, you list the possible outcomes, attach a probability to each, attach a value ("utility") to each, multiply value by probability, and sum. The option with the highest total wins. Two ideas make this powerful. First, it cleanly separates *how likely* something is from *how much it matters* and then recombines them by one consistent rule. Second, the value function is usually curved, not straight: an extra dollar matters less to a rich person than to a poor one. That curve is why most people prefer a guaranteed $50 over a 50-50 shot at $0 or $100, even though the average money is the same. The shape of the curve encodes how much you dislike risk.

Expected utility is a structural rule for valuing uncertain prospects: weight the utility of each possible outcome by its probability and sum. This collapses an entire distribution of futures into a single comparable scalar that ranks options under risk. Bernoulli (1738) introduced the core move when resolving the St. Petersburg paradox, proposing that people value the *logarithm* of wealth, so a gamble with infinite expected dollars still commands only a finite price. Von Neumann and Morgenstern (1944) put the framework on rigorous axiomatic footing, showing that any agent whose preferences over risky prospects satisfy a short list of consistency axioms (completeness, transitivity, continuity, independence) must behave *as if* maximizing the expectation of some utility function — the utility function is recovered from preferences, not imposed. The leverage comes from the *curvature* of the utility function: concavity (each extra unit of the good worth less than the last) automatically yields risk aversion, because the upside is valued less steeply than the downside is penalized. Expected utility is thus not merely an averaging operation but a valuation discipline in which an agent's attitude toward risk is a readable property of the utility curve.

Broad Use¶

Economics / decision theory: the von Neumann–Morgenstern criterion for rational choice under risk.
Finance: pricing and portfolio choice trading expected return against risk via a utility of wealth.
Artificial intelligence: expected-value maximization in decision-theoretic planning and reinforcement learning (expected return).
Biology / ecology (non-obvious): risk-sensitive foraging, where animals choose between variable food patches as if maximizing expected fitness, not expected calories.
Engineering / reliability: expected-cost decisions weighting failure severities by their probabilities.
Public policy: cost-benefit analysis under uncertainty using expected outcomes.

Clarity¶

Naming expected utility lets practitioners see that rational choice under uncertainty separates two ingredients — how likely and how much it matters — and combines them multiplicatively. It clarifies why a rational agent may reject a positive-expected-money bet (concave utility) and exposes when a decision rule departs from this benchmark.

Manages Complexity¶

It reduces a branching tree of uncertain futures, each with its own payoff, to one scalar score per option, making otherwise incomparable risky alternatives directly rankable. This bounding move is what turns "what might happen?" into "which option scores highest?"

Abstract Reasoning¶

Recognizing the structure supports inferences about risk attitude (curvature of the value function encodes aversion or seeking), about why diversification and insurance are rational, and about systematic human deviations (prospect theory) measured precisely as departures from the expected-utility baseline.

Knowledge Transfer¶

The decision-theoretic template transfers from financial portfolio choice to AI agents weighting action outcomes by predicted reward, to evolutionary models where genotypes "hedge" against environmental variance, to medical decision analysis weighting treatment outcomes by probability and patient-valued utility.

Example¶

Offered a 50/50 chance at $0 or $100 versus a sure $40, a risk-averse agent with concave utility takes the sure $40 even though the gamble's expected money is $50, because the expected utility of the certain amount exceeds that of the gamble. The same calculus governs a forager choosing a reliable patch over a variable richer one and an RL agent selecting the action with the highest expected return.

Relationships to Other Primes¶

Parents (3) — more general patterns this builds on

Expected Utility presupposes Preference — Expected utility presupposes preference because probability-weighted aggregation requires a prior utility function ranking outcomes.
Expected Utility is part of Probability — Expected utility is a constituent piece of probability reasoning; it provides the probability-weighted aggregation of value over outcomes.
Expected Utility is a decomposition of Aggregation — Expected utility is the specific shape aggregation takes when uncertain outcomes are collapsed into one scalar by probability-weighted summation of a utility function.

Children (1) — more specific cases that build on this

Risk Aversion presupposes Expected Utility — Risk aversion presupposes expected utility because the certainty-over-gamble preference is formally defined as concavity of the expected-utility value function.

Path to root: Expected Utility → Preference

Not to Be Confused With¶

Expected utility is not marginal_utility, the change in utility from one more unit of a good; expected utility aggregates a value function across uncertain outcomes using probabilities. It is not optionality, the asymmetric value of a right without obligation; expected utility scores any prospect, asymmetric or not. Unsure flag: the candidate sits near the line between a structural pattern (probability-weighted aggregation of value, which recurs broadly) and a decision-theoretic criterion/method; the curator should judge whether the catalog admits such normative-formal patterns as primes.