Conditional probability measures the likelihood of an event occurring when another event is already known to have occurred. It allows reasoning to be updated when additional information becomes available. For events A and B, the conditional probability of A given B is defined using the probability of their intersection divided by the probability of B, provided that B has a nonzero probability. Bayes' theorem follows from the relationship between conditional probabilities and provides a method for reversing the direction of conditioning. It combines a prior probability, a likelihood, and evidence to calculate a posterior probability. The prior represents the initial belief about a hypothesis, the likelihood describes how probable the observed evidence is under that hypothesis, and the evidence term accounts for the overall probability of the observation. Bayes' theorem is important in AI because systems frequently need to update predictions when new observations arrive. It is used in Bayesian classification, medical decision support, spam filtering, fault diagnosis, and probabilistic inference. Correct application requires carefully defining the events and interpreting conditional probabilities in the correct direction.
Conditional Probability and Bayes' Theorem
An explanation of conditional probability and Bayes' theorem for updating event probabilities when new information becomes available.
P(A | B) = [P(B | A) × P(A)] / P(B)
Explanation
Variables
| Symbol | Meaning |
|---|---|
P(A | B) | The probability of A given that B has occurred; this is the posterior probability in a Bayesian update. |
P(B | A) | The probability of observing B when A is true; this is the likelihood. |
P(A) | The prior probability of A before observing B. |
P(B) | The probability of observing B, also called the evidence. |
Example or application
Suppose a medical condition affects 1% of a population. A test has 90% sensitivity and a 5% false-positive rate. For a randomly selected person who tests positive, Bayes' theorem gives P(condition | positive) = (0.90 × 0.01) / [(0.90 × 0.01) + (0.05 × 0.99)] ≈ 0.154. Thus, the probability is approximately 15.4%, illustrating why the prior rate and false positives matter.
Text
Conditional probability evaluates an event using the assumption that another event has occurred. Bayes' theorem uses conditional probabilities to update an initial probability after observing evidence. It relates prior probability, likelihood, evidence, and posterior probability. This makes it useful for AI systems that revise predictions when new information becomes available.
Key Points
Conditional probability considers an event under the condition that another event has occurred.
The conditioning event must have a nonzero probability.
Bayes' theorem reverses the direction of conditional probability.
The prior represents an initial probability before new evidence.
The likelihood measures how probable the evidence is under a hypothesis.
The evidence term accounts for the probability of the observed information.
The posterior is the updated probability after considering evidence.
Bayesian reasoning is widely used in probabilistic AI and machine learning.
Applications
Naive Bayes classification
Email spam detection
Medical test interpretation
Fault diagnosis
Bayesian networks
Probabilistic reasoning
Risk assessment
Predictive decision support
Learning Outcomes
Define conditional probability.
Explain the relationship between joint and conditional probability.
Identify prior, likelihood, evidence, and posterior probabilities.
Apply Bayes' theorem to a numerical example.
Explain why prior probabilities matter when interpreting test results.
Practice Questions
What is conditional probability?
State Bayes' theorem.
What is the difference between prior and posterior probability?
What does likelihood represent in Bayesian reasoning?
Why can a positive medical test result still correspond to a relatively low probability of disease?
Related Concepts
Probability Theory for AI
Statistics and Probability Distributions
Bayesian Inference
Naive Bayes Classifier
Probabilistic Graphical Models
Common Mistakes
Reversing P(A | B) and P(B | A).
Ignoring the prior probability when applying Bayes' theorem.
Forgetting that the evidence term must account for all relevant ways the observation can occur.
Treating a likelihood as if it were automatically a posterior probability.
Applying the conditional probability formula when the conditioning event has zero probability.
Sources and verification
Sources are shown as provided in the project data. Not every source has yet been independently verified. Please report errors or missing citations.
- Bayes, T. An Essay towards Solving a Problem in the Doctrine of Chances.
- Ross, S. M. A First Course in Probability.
- Murphy, K. P. Probabilistic Machine Learning: An Introduction.
Suggest a correction or source · More in Artificial Intelligence & Machine Learning