Knowledge Tracing Simulator illustration

Knowledge Tracing Simulator

Bayesian Knowledge Tracing is how intelligent tutoring systems (Khan Academy-style mastery learning, ASSISTments, cognitive tutors) decide whether you have learned a skill. A hidden state — known or not known — evolves with a learning probability p(T); observed answers are noisy through guessing p(G) and slipping p(S). Watch the system update its mastery estimate after every answer; a step-by-step panel plugs the percentages into Bayes' rule so you can see why a correct guess barely moves the estimate and a slip can crash it. Tune the four parameters to understand why some tutors ask for three correct answers in a row before moving on.

Runs 100% in your browser — simulations are computed locally on your device.

Notes

  • The posterior update is plain Bayes' rule: a correct answer raises the mastery estimate more when guessing is unlikely; a slip can wrongly crater it. The panel lists prior → likelihoods → evidence → posterior → learning transition for every attempt.
  • With high p(G) (easy multiple choice), mastery estimates rise slowly — the system cannot tell knowledge from luck, which is why good items matter more than good algorithms.
  • Runs 100% in your browser — nothing you type or practice leaves your device.

Explore the idea

This interactive demo needs JavaScript.