Can You Explain It With the Tab Closed?

Look away from the screen and explain the last thing an AI taught you, out loud. Say what the idea is and give one example.

What comes out may look like an explanation with nothing inside it. Picture a course finished in half the time, with AI summarising every module and the certificate already in your inbox. Then a colleague asks for the main idea, and the words aren't there.

Going through something and learning it feel the same from the inside, and AI has made the first one cheap.

Eva Keiffenheim writes about learning in the Lifelong Learning Club. She has a rule that settles most of the argument about whether AI is good for learning. AI helps when it makes you recall, and it harms when it thinks for you. The rule is hers.

The reason is well known from how memory works. You keep what you pull back out of your own head. A summary you read is pleasant, but it's gone in a day. A question you had to answer from memory, badly, and then corrected, stays.

AI can hand you either one, and which you get depends entirely on what you ask for.

Keiffenheim has a second idea, a prompt she calls neutral ground. Before you learn from AI, ask what ideas and slant it brings to the subject. Then you learn the subject, and you can also see the AI's angle on it.

A confident explanation hides its slant, so it's worth asking, though once per topic is enough.

I'm about to learn about [topic] from you. Before you explain anything, answer three questions. What way of seeing the subject will your explanation lean on? What do people in this field disagree about? And what would someone who disagrees with the usual view tell me to watch for? Keep it short. Then I'll ask my questions.

Now learn the topic however you like. Ask questions, follow the side paths and enjoy this part.

Then close the tab and explain the topic without looking. Say it out loud to a person or to the wall, or write it down. Aim for one paragraph that says what the idea is and why it matters, with one example.

Your paragraph will be worse than you expect, and that's the recall at work. Every gap you hit is a thing you didn't learn, and now you know where the gap sits.

Reopen the tab. Ask the AI to mark your explanation, and to leave your words as they are.

Here's my explanation of [topic] from memory. [paste or type what you said] Don't rewrite it. Tell me what's wrong, what's missing, and what I've got right. Then ask me one question that would test whether I understand the thing I got wrong.

Answer its question, then close the tab again. Tomorrow, explain the topic once more without looking.

Would you rather the AI quizzed you from the start? Then ask "Don't explain this. Ask me questions about it, one at a time, and only explain when I get something wrong."

That's harder work, and it sticks better.

Most of what you ask AI in a week is looking things up, and looking things up doesn't need the test. If you want the VAT rate on books, don't waste a good test on that.

Save the test for things you'd want in your head a year from now, where a chat history won't do.

Think of a patient tutor that's always there and explains anything in whatever way you need. That's a gift, and the people it helps most are the ones who'd never have had a tutor at all.

Trouble starts when you stop at the explanation. So keep the tutor, and add the recall. Whatever paragraph you managed with the tab closed, gaps included, is the part you own.

AI helps when it makes you recall, and it harms when it thinks for you.

Credit to Eva Keiffenheim, from the Lifelong Learning Club (2025). The line "it helps when it makes you recall, harms when it thinks for you" and the neutral-ground prompt are hers. Read her for the full seven principles.