Learning With AI: Put a Recall Check Before the Next Lesson

You ask an AI assistant to teach you how to read an online shop report. It explains the difference between an order and an item, then moves straight on to average order value. By the fourth explanation, you can follow every sentence. Ten minutes later, you cannot decide what to count in an order containing three items. The conversation felt easy to follow because the explanation was always in front of you. You never had to decide what to do with it.

Try adding a deliberate pause before the next section: hide the explanation and answer one question about what you just learnt. In this example, count the orders and items in a tiny fictional basket. The answer will help you decide whether to move on or revisit the distinction. It also changes the structure of the lesson: you get a turn to think before the assistant gives you more to read.

What researchers tested between lecture segments

In a 2013 paper in PNAS, Karl Szpunar, Novall Khan and Daniel Schacter reported two experiments using a short recorded introductory statistics lecture. The researchers split it into four segments and compared students who took brief memory tests between segments with students who did not take those interim tests. The second experiment also included a group that reread questions with answers instead of attempting them during the early breaks.

The groups given interim tests reported less mind wandering and showed better learning in the study’s assessments. What makes the work relevant to an AI tutoring session is where the questions appeared: participants encountered them while the lesson was still in progress. The study did not use chatbots, examine every kind of learner or establish how often a person should interrupt a lesson. It gives you reason to try a brief recall check, rather than a schedule every lesson has to follow.

Make the pause do one job

Suppose the shop owner studies a report with two fictional orders: one contains a notebook and a pen; the other contains two cards. Ask a question that checks one distinction: how many orders were placed, and how many items were sold? The answer is two orders and four items. If you answer four orders, the question has identified a specific misunderstanding. You can revisit what counts as an order before the lesson asks you to divide revenue by that number.

This is more useful than asking whether you understood the explanation. It also gives you a better check than a question whose answer is almost visible in the wording. Keep the source and the answer out of view until you commit to a response. If your first answer is correct but the reasoning is shaky, try a second basket with different numbers. If your first answer is wrong, compare your counting rule with the verified example and try again. You are choosing what to study next. Keep a note of the first answer and the rule you used. If the same confusion comes back later, the note helps you find a better example rather than repeating the same explanation word for word.

A good place to stop is after a complete idea. Interrupting halfway through a demonstration may leave you trying to reconstruct something you have not yet seen. Waiting until the end of a long session can make it harder to identify where you lost the thread. Let the subject set the pause: after a distinction, a worked example or an instruction you will need to use in the next segment.

Use the assistant to manage the turn-taking

• Set a stopping point. Ask for one idea or procedure at a time. Tell the assistant to stop after a short question and wait for your reply without giving the answer.
• Check the question against the material. For a course or work procedure, use an approved example or a reliable source. If the question depends on a fact the lesson did not establish, revise it rather than treating a wrong answer as a learning failure.
• Let the answer decide the next step. If you can explain the rule, move on. If you have the right number for the wrong reason, work through a fresh example. If you are stuck, ask for one hint and then answer again.
• Keep the next practice task independent. When the lesson ends, try one small example by yourself. Otherwise a useful pause can turn into another exchange in which the assistant does most of the work.

There is a trade-off. Too many questions can make a coherent lesson feel fragmented. Some material needs an uninterrupted first pass, and someone learning a completely unfamiliar task may need to see a worked example before a fair question is possible. MaryChuks.com’s guide to checking how much help it takes to regain a skill addresses a different decision: what to practise when you return to the task later. A pause during the lesson tells you what you can use now; it cannot tell you what you will remember next week.

For your next AI lesson, decide on one idea that the rest of the session depends on. Make the assistant wait after asking about it. If your answer shows you are ready, carry on. If it exposes a gap, you have found it while the lesson is still short enough to repair.

Source

• Karl K. Szpunar, Novall Y. Khan and Daniel L. Schacter, Interpolated memory tests reduce mind wandering and improve learning of online lectures, PNAS (April 2013).