The History of AI

Bayes's essay

Era 1 · The foundations (1763–1942) › Probability and statistics

Lineage
Data and learning
Kind
Publication

Who

Royal Society

What happened

On December 23, 1763, "An Essay towards solving a Problem in the Doctrine of Chances," by Thomas Bayes, who had died two years earlier, is read at the Royal Society. It was prepared by a friend, Richard Price.

Why it mattered

It works out how to go from what is observed to the probability of what cannot be seen. Its propositions contain what is now called Bayes's theorem, the basis of inference that updates a belief with every new piece of data.

The full story

Thomas Bayes, a Presbyterian minister and amateur mathematician, died in 1761 without publishing the work that made the name famous. The family handed the papers to Richard Price, another minister with mathematical training, who spent two years editing them, wrote an introduction, and chose one of the two solutions Bayes proposed. The essay was read at the Royal Society on December 23, 1763, and appeared in volume 53 of its Philosophical Transactions.

The problem runs the usual one in reverse. Classical probability starts from a known cause, such as a fair die, and calculates what will be observed. Bayes starts from what has been observed, how many times an event has happened and how many times it has not, and asks what can be said about the unknown probability behind it. To set the problem up, the essay imagines balls thrown onto a square table: where the first one lands fixes a probability nobody sees, and the rest fall to one side of it or the other.

The propositions contain the relation between prior and posterior probability that now bears Bayes's name, though the essay does not single it out. Price saw in it an argument for the existence of God.

Pierre-Simon Laplace arrived at the same result in 1774 without knowing this work. Two hundred years later, the idea of updating a belief with each piece of data underlies Bayesian statistics and many machine learning methods.

Under the hood

In modern notation, the theorem relates the probability of a hypothesis H after seeing some data D to the probability it had before: P(H|D) = P(D|H) · P(H) / P(D). P(H) is the prior probability, P(D|H) the probability of the data if the hypothesis is true, and P(H|D) the posterior.

Bayes's problem is a specific case: after p successes and q failures, what is the probability that the unknown probability of success lies between two values? With a uniform prior, the posterior is what is now called the beta distribution.

Common misconceptions

  • What people often say: Bayes wrote down the formula that now bears that name.

    In fact: The essay solves a specific problem with geometric reasoning. The general formula, in the form taught today, is Laplace's.

Leads to

Sources

Original source

How to cite

“Bayes's essay.” The History of AI, revised November 13, 2026. https://thehistoryofai.com/en/milestone/bayes-essay

This entry was last reviewed on November 13, 2026.

Open in the interactive timeline