Adaptive Learning, Explained Simply
"Adaptive" is one of the most over-used words in education apps. Usually it just means the app gets a little faster or slower. In Amal it means something more concrete: the knowledge graph builds a different path for every child, updated as they go. Here is how — with the jargon translated into plain words.
Step 1: What is my child ready for next?
Amal keeps a private record of what your child already knows. Match that record against the graph's prerequisite map, and you can answer one precise question: which concepts has this child mastered all the prerequisites for, but not yet learned?
That set of "ready right now" concepts is called the learner frontier. It is not too easy (things they've already mastered) and not too hard (things whose prerequisites aren't in place yet). It's the just-right edge of what they can learn today.
Step 2: Build a tree just for this child
Amal turns that frontier into a learning tree — the subjects and lessons your child sees when they open the app. The key part: the tree is rebuilt for each child from their own progress.
| Layla (just starting) | Omar (a few weeks in) | |
|---|---|---|
| Mastered | Letter shapes | Letters + first words |
| Frontier (next up) | Letter sounds | Joining letters, short words |
| Tree shows | Sound practice | Word-building lessons |
Two children, the same graph, two genuinely different trees. Nobody hand-assigns these paths; they fall out of each child's own mastery.
Step 3: Bring it back before it's forgotten
Learning something once isn't the same as remembering it. Amal uses a memory model called HLR (Half-Life Regression) — from the same family of spaced-repetition science that Duolingo helped popularise. In plain words: it estimates how likely your child is to still remember each concept, and schedules a review just before they'd otherwise forget it. Review too early and you waste time; too late and it's gone. HLR aims for the sweet spot.
Step 4: Fresh practice, not a fixed deck
The actual lesson — what Amal calls a content byte — is generated from the concept, not pulled from a fixed deck of cards. So practice stays fresh: a child revisiting the letter ب doesn't see the exact same screen every time. Each byte is a few short steps stitched together — for a letter, that might be see it → recognise it → say it.
How we make sure it actually adapts
Here's an honesty beat, because "adaptive" is easy to fake. Early in development, our adaptivity signal was degenerate: if a child could tap through a "just say it" step, everyone came out looking equally "mastered." No real differentiation — the exact thing adaptive learning is supposed to avoid.
We fixed it by being strict about which steps count. Steps that merely expose a child to a concept grant a little credit, capped. Only the scored steps — where the child actually has to recognise or apply something — move real mastery. The lesson we took away: an adaptive system has to prove it produces different outcomes for different children, not just that it recorded something. That's the bar we hold it to now.
The rest of the series
- How Alphazed Teaches Arabic With One Knowledge Graph — what a knowledge graph is.
- One Graph, Every Lesson: Apps, Books, and Printables — one concept, many surfaces.
- Adaptive Learning: A Knowledge Graph Per Child — you are here.
- Why a Knowledge Graph Is Good for Your Child — the benefits for kids.
- For Parents: See Exactly What Your Child Knows — the map you can see.
Curious to watch your own child's path unfold? Start with Amal.