Skip to main content

Hints, Retrieval and Transfer

What this concept is​

An assisted result shows what you can do with that assistance. Independent retrieval shows what you can explain without the answer in view. Transfer checks whether you can use the idea when the example changes. These are different observations.

Why it matters here​

An AI tutor can reduce the difficulty of the current question so much that you never practice the skill you intended to learn. Use assistance to expose the next useful step, then test yourself on a fresh case.

Concrete example​

Task: fit two 25-minute sessions and a 10-minute review into an available block. First calculate the total yourself. A helpful hint asks, “Which activities must be counted, and what units are they in?” A complete answer may be useful after an attempt, but copying it is not evidence that you can plan.

After feedback, close the solution and solve a new case: three 15-minute sessions plus a 10-minute review in 60 minutes. Explain the five minutes remaining.

Common confusion / misconception​

Recognizing an explanation while reading is easier than producing it. Refusing all assistance is not the objective either. The learning goal determines when to use a hint and when an independent attempt is necessary.

How to use it​

Record a prediction, attempt, chosen hint level and revision. Ask one question at a time. Preserve what changed in your reasoning. Later, close the transcript and solve a changed case. If you cannot, reduce the task, repair the missing idea and try another case rather than claiming mastery.

Check yourself​

A perfect assisted worksheet and a failed independent variant mean what? The assistance helped that worksheet; independent competence is still unestablished. It is a repair signal, not a judgment about ability.

Mini drill or application​

Use the offline tutor fixture in the guided lab. Write your own attempt before revealing the hint. Record the fresh-case result separately. A live AI response is optional and must be labeled if used.

Read this only if stuck​

Bastani and colleagues' classroom experiment distinguishes assisted task performance from later unassisted performance in high-school mathematics. It motivates a check here; it does not establish an identical effect for adult computing learners or all AI tools.