Bayes and probability: how to reason under uncertainty

What is it?

Probabilistic models do not give one certain answer; they estimate how likely different outcomes are. The Bayesian approach starts with a prior belief and updates it as new evidence arrives. That lets a system reason even when it does not know everything.

An everyday example

A spam filter estimates how likely a message is to be spam from the words in it. When you mark a message as spam, the filter updates its estimate.

Why does it matter?

A good estimate of uncertainty tells you when an answer can be trusted and when to check further. That is crucial in medicine, finance and any important decision.

A common misconception

That a 90 percent probability means 90 percent certainty that the model is right. The number is only worth as much as the data and assumptions behind it.

In more detail

What is a probabilistic model?

It describes the possible outcomes and how likely each is, given the information and assumptions.

What does the Bayesian approach mean?

Updating your estimate of the odds when new data arrives.

Does 90% probability mean 90% certainty?

Only if the forecast is well calibrated and refers to a correctly described task.

Why is it useful?

For medical tests, risk forecasting, spotting faults and making decisions under uncertainty.

Remember: A probability figure is only useful if we know how it was produced.