All guidesTeach-back
Teach-back
AI student: Explain it to an AI student — teaching is the strongest way to learn.
Teach-back flips the roles: the AI becomes a student who does not quite get it, and you become the teacher. Explaining aloud is where shallow understanding shows — and where it gets fixed.
What it is
On any article or video lesson, start a teach-back conversation. The AI plays a learner studying this exact lesson: it asks you to explain, restates what it understood, and points at the parts that lose it — grounded only in this lesson’s content.
When you have taught your best, run the evaluation. You get a structured assessment: which points you explained correctly, any misconceptions in your explanation with the accurate statement, and overall advice.
Why — the pedagogical rationale
The protégé effect is well documented: preparing to teach, and teaching itself, force you to retrieve knowledge, organize it coherently, and confront the gaps that passive reading hides. Explaining to a genuine questioner is harder — and more effective — than summarizing to yourself.
The evaluation then closes the loop into your twin: concepts you explained correctly earn mastery credit (a strong verbal-recall signal, though deliberately weaker than a correct assessment), and anything the evaluation catches becomes a suspected misconception your tutor will keep an eye on.
How we achieve it
The simulated student is synthesized server-side from your course’s teaching style, with retrieval scoped to the lesson being taught back — it stays on topic and grounded. The evaluation judges your explanation against the concepts extracted for this lesson, so its outcomes are specific, not vibes.
Like every AI feature, the evaluation is quota-gated and logged; the twin writes it produces are exactly the ones you can see on your twin page.
Where to find it
Open an article or video lesson and choose "Teach-back". The sidebar walks you through the conversation and the evaluation.