Jaynes-Cox Theorem: Why Probability is Rational Logic?

Brief Introduction

The Jaynes-Cox Theorem (Jaynes-Cox Theorem), commonly known as Cox's Theorem, is a cornerstone in the intersection of probability theory and logic. Proposed by R.T. Cox in 1946, it was later vigorously popularized by E.T. Jaynes in his work Probability Theory: The Logic of Science. It proves a startling conclusion: any rational inference system that satisfies consistency, continuity, and correspondence with common sense is mathematically equivalent to probability theory. This means that probability is not merely a tool for handling random events, but the only reasonable way for humans to perform logical inference under uncertainty.

Core Knowledge Points

The core of this theorem lies in the axiomatic definition of "rational inference." Cox and Jaynes proposed three basic assumptions:

1. Continuity: The confidence of inference can be represented by real numbers, and changes are continuous.
2. Associativity and Combination: The order of joint inference of multiple propositions should not affect the final result, meaning $(A|B|C)$ should equal some functional form of $(A|C|B)$.
3. Correspondence with Common Sense: The inference system should align with our intuitive understanding of "true," "false," and "irrelevant."

Based on these seemingly mild conditions, mathematical derivation shows that the function describing uncertainty must satisfy the probability multiplication rule and the probability addition rule. Therefore, the Bayesian updating formula is no longer an artificially chosen technique, but a necessary requirement of rational thinking.

Relation to the Content of "Bayesian Games"

In the context of Bayesian Games, the Jaynes-Cox Theorem provides a deep philosophical justification for games with incomplete information in game theory.

The game participants discussed in the book often face information asymmetry and need to continuously update their beliefs about opponents' strategies. Traditional views might consider Bayesian updating merely a "method," but this theorem points out that if a rational agent wishes to avoid logical contradictions (such as a Dutch Book), it must follow Bayesian rules. In game scenarios, this implies:
* Necessity of Belief Updating: Participants cannot arbitrarily change probability estimates; they must strictly calculate posterior probabilities based on new evidence.
* Rational Basis of Strategy: Strategy choices based on probability (such as mixed strategy Nash equilibrium) are not just mathematical solutions, but manifestations of logical consistency.

This theorem elevates game theory from mere mathematical calculation to the height of "rational logic," explaining why Bayesian agents have advantages in long-term evolution.

Summary

The Jaynes-Cox Theorem reveals the essence that Probability is Logic. Whether in scientific research, artificial intelligence decision-making, or strategic interactions in Bayesian Games, it establishes probability theory as the sole standard for inference under uncertainty. Understanding this theorem helps us gain a deeper appreciation of the universality and necessity of Bayesian methods.

Keywords: Jaynes-Cox Theorem, Probability Theory, Bayesian Games, Rational Inference, Logic of Uncertainty, E.T. Jaynes, R.T. Cox