Bayesian Updating
Bayesian updating is the process of revising the probability of a hypothesis based on new evidence, often quantified as a posterior probability.
THE IDEA
Bayesian Updating
Bayesian updating involves using Bayes' theorem to adjust the probability of a hypothesis as new evidence becomes available. The principle is widely applied in probability, statistics, psychology, and decision-making to model rational belief revision.
Example from Text
"EM: [The saboteur is] one of us . . . ? SPOCK: Approximately 82.5 percent in favor of the possibility. Indeed, the saboteur is in their group."
This exchange gives an 82.5% estimate, but it does not show a prior probability, new evidence, or a Bayesian calculation.
When to use it
Assessing the likelihood that a team member is responsible for a mistake based on prior performance and new clues.
Evaluating the probability that a patient has a rare disease after receiving a positive test result.
Updating the chance of project success as new information about risks or milestones emerges.
Estimating the reliability of a witness testimony by combining prior credibility with new contradictions.
FROM THEORY TO EVERYDAY LIFE
In practice
Saboteur Group Estimate
Spock gives an 82.5% estimate that the saboteur is in the group. The exchange does not show how he arrived at it.
Medical Diagnosis
In a hypothetical example, a doctor updates the probability of a disease from 1% to 85% after a positive test result, using Bayesian reasoning to combine prior prevalence with test accuracy.
A MOMENT TO REFLECT
Put it to work.
How would you estimate the probability that a specific person in a group is a saboteur, given a base rate of saboteur presence?
In what real-life situations might you need to update a belief based on new evidence, and how could Bayesian reasoning improve your estimate?
Why might people often neglect prior probabilities when making judgments, as seen in the base rate fallacy?
Often confused with
CONTINUE THE CONVERSATION
Source texts
- The Scout Mindset (Julia Galef)