Mathematical Axioms of Expected Utility Theory

Expected utility theory is a cornerstone of modern game theory and decision theory, providing a framework for analyzing choices under uncertainty. This article explores the core mathematical axioms—completeness, transitivity, continuity, and independence—established by John von Neumann and Oskar Morgenstern, which guarantee that a rational decision-maker’s preferences can be represented by a utility function.

The Von Neumann-Morgenstern Utility Theorem

To model decisions under risk, game theory relies on the Von Neumann-Morgenstern (VNM) formulation. This framework assumes a decision-maker faces choices among “lotteries” (probability distributions over a set of outcomes). For a decision-maker’s preferences to be mathematically represented by an expected utility function, their choices must satisfy four foundational axioms.

1. Completeness

The axiom of completeness assumes that an individual has well-defined preferences and can always compare any two alternatives.

For any two lotteries \(A\) and \(B\), exactly one of the following must hold: * \(A \succ B\) (A is strictly preferred to B) * \(B \succ A\) (B is strictly preferred to A) * \(A \sim B\) (The decision-maker is indifferent between A and B)

This rules out the possibility that a decision-maker is unable to make a choice or compare options.

2. Transitivity

Transitivity ensures that a decision-maker’s preferences are internally consistent.

If a decision-maker prefers lottery \(A\) to lottery \(B\), and prefers lottery \(B\) to lottery \(C\), then they must prefer lottery \(A\) to lottery \(C\): \[\text{If } A \succeq B \text{ and } B \succeq C, \text{ then } A \succeq C\]

Without transitivity, a decision-maker could hold cyclical preferences (preferring A to B, B to C, and C to A), making rational choice impossible and leaving them vulnerable to being used as a “money pump” in economic transactions.

3. Continuity

The continuity axiom states that there is no infinitely good or infinitely bad outcome. It establishes a middle ground between extreme choices using probabilities.

If a decision-maker prefers \(A\) to \(B\), and \(B\) to \(C\), there must exist a probability \(p \in (0, 1)\) such that the decision-maker is indifferent between receiving the middle option \(B\) for sure, or taking a lottery that yields \(A\) with probability \(p\) and \(C\) with probability \(1-p\): \[pA + (1-p)C \sim B\]

This axiom ensures that preferences do not jump abruptly and rules out lexicographic preferences (where one dimension of a choice completely overrides another, regardless of probability).

4. Independence

The independence axiom asserts that if we mix two lotteries with a third common lottery, the relative preference between the original two lotteries does not change.

If \(A\) is preferred to \(B\), then for any third lottery \(C\) and any probability \(p \in (0, 1]\): \[pA + (1-p)C \succeq pB + (1-p)C\]

This means that a decision-maker’s preference between \(A\) and \(B\) should be independent of the presence of an irrelevant alternative \(C\), assuming both are scaled by the exact same probability.

The Expected Utility Representation

When a decision-maker’s preferences satisfy these four axioms, the VNM Utility Theorem proves that there exists a real-valued utility function \(u\) that assigns a number to each outcome. The utility of any lottery \(L\) with outcomes \(x_i\) occurring with probabilities \(p_i\) is simply the expected value of the utilities of those outcomes:

\[U(L) = \sum p_i u(x_i)\]

In game theory, this allows complex tactical interactions to be modeled numerically, as players seek to maximize their expected utility based on their opponents’ anticipated actions.