Methods · Assessment & diagnosis · Short guide
H4
Mastery threshold checking
Keep going until a defined performance criterion is met, not until the clock or the syllabus says you are done.
When to use it
- You finished the chapter calendar but cannot say what “known” means for this skill.
- You want a gate before moving on in a cumulative sequence.
- You keep adding new content instead of proving the last skill.
What to ask AI
Ask it to propose a mastery threshold for the skill that you can observe, include accuracy plus one transfer or explanation element.
Then ask it to administer a check without hints. If you fail, remediate only the miss and recheck.
Refuse to move on until the threshold is met twice, not once by luck.
How it helps
Time-based progression advances the calendar, not competence.
Threshold checking asks what performance counts as mastery here, then measures it before unlock.
A good threshold is observable: for example four of five mixed problems plus explaining one “why.”
This is the assessment-framed cousin of exiting a practice session only when you hit criterion.
How to run it
- Define the threshold in observable terms before you practise.
- Assess under realistic constraints, less help than study mode.
- Pass → progress. Fail → remediate only the miss, then recheck.
- Log attempts so you can see if the threshold is unrealistic or if you keep almost-passing.
Example exchange
Skill I'm aiming for: interpreting p-values. Propose a mastery threshold, then test me without hints.
Threshold: four out of five scenario questions correct, plus two sentences on what a p-value is not. Ready for the check?
Ready. Here are my answers.
You hit three of five, and the “what it is not” sentence slipped. We'll remediate only the false-positive confusion, then recheck the same threshold, not a new one.
Copyable prompt
Skill: [ ]. Propose a mastery threshold that is observable and includes at
least one transfer/explanation element. Then administer a check without hints.
If I fail, remediate only the miss, then recheck.
The Tell
The syllabus is complete and the thresholds were never measured.
Finishing time is not the same as meeting a criterion.
One lucky pass also fails the method, require a clean recheck.
Principle evidence
Strength of the underlying learning idea, not a claim about AI products.
The underlying learning idea is rated moderate. Mastery thresholds follow mastery-learning logic (Bloom tradition): advance on criterion, not seat time. Credibility rises when criteria are well defined.
AI delivery evidence
Whether an AI tutor delivers this method well is a separate question.
AI gatekeeping of “mastery” is speculative. You own the threshold definition and the decision to unlock the next topic.