SoulTraceSoulTrace
Adaptive Testing

Bayesian Inference

Also called: Bayesian updating, Bayes inference

Bayesian inference is a method for updating probabilities when new evidence arrives. An assessment begins with prior probabilities over possible trait values or profiles, evaluates how likely each answer would be under those possibilities, and produces updated posterior probabilities. Repeating this process makes uncertainty explicit as the test progresses.

Reviewed July 14, 2026 · 2 min read

Table of Contents

When faced with a complex decision, I prioritize a methodical approach over intuitive leaps.

Neutral
Strongly DisagreeStrongly Agree
/en/new-test?utm_source=glossary&utm_medium=internal&utm_campaign=glossary-assessment&utm_content=bayesian-inference-after-definition-cta-assessment-test

Prior, likelihood, and posterior

Bayesian updating combines three pieces:

  • the prior, which represents uncertainty before the new answer;
  • the likelihood, which describes how probable that answer is under each possible profile;
  • the posterior, the updated probability distribution after combining them.

The posterior becomes the prior for the next question. Evidence accumulates rather than restarting after every response.

A personality-assessment example

Suppose several profiles remain plausible. A response favoring careful planning is more expected under some profiles than others. Their probabilities rise relative to profiles where that response is less expected. One answer rarely determines the result because the model combines the entire response pattern.

If the answer is equally likely under every remaining profile, it changes little. That is why an adaptive system tries to select questions whose possible answers would separate the current possibilities.

Why distributions matter

A single winning label hides whether the model is nearly certain or split between close alternatives. A posterior distribution retains that distinction. It can guide question selection, define stopping rules, and support honest result language.

The prior is not arbitrary guesswork

Priors can be broad, based on relevant population data, or structured by the model. Their influence usually decreases as informative responses accumulate, but they should still be documented and tested. Bayesian mathematics does not rescue weak assumptions or biased evidence; it makes those assumptions part of an explicit model.

Go deeper: Bayesian inference inside Soultrace

Sources

Related terms

Probability DistributionA probability distribution assigns probabilities or density across possible outcomes. In personality testing, it can represent uncertainty over trait values, profiles, or archetypes after the available answers. A concentrated distribution indicates that the model favors a narrow set of possibilities; a spread-out distribution shows that several interpretations remain plausible.Information GainInformation gain measures how much an observation is expected to reduce uncertainty. In adaptive testing, the system can predict the possible answers to each eligible question, estimate the posterior uncertainty after each answer, and choose the question with the greatest average reduction. The actual gain is known only after a response is observed.EntropyEntropy is a mathematical summary of uncertainty in a probability distribution. In an adaptive personality assessment, entropy is higher when probability is spread across several plausible results and lower when it is concentrated on fewer possibilities. A question can be valuable when its expected answers would reduce entropy.