What adaptive learning means in practice
Most ‘adaptive’ KS2 apps move difficulty up or down based on whether a child got the previous answer right. That's a single signal — useful, but blunt. Northstar treats adaptation as a planning problem: the engine knows the curriculum, the prerequisite chains between skills, the misconceptions that commonly surface at each year, and the per-child cognition profile that summarises what's been seen and how it landed. The result is that two children working on the same Year 5 fractions topic at 10 a.m. see different questions. One who slipped on comparing fractions yesterday gets a calibration question first; another who's secure there moves straight to fractions of amounts. Neither gets bored, neither stalls.
Inside Northstar's adaptive engine
Four parts. None are visible to the child — they only see the next question. But each one materially changes what's chosen.
Cognition profile
Per-child, per-subject summary of mastery, drift, and prerequisite gaps. Updated after every session. Used by every other layer below.
Verifier-first AI
When Live AI proposes a question, a closed-rule gate re-derives the correct answer from a structured contract and rejects anything it can't verify. AI is never trusted for correctness — the gate is.
Mastery loop
Wrong answers route to a calibration question on the underlying misconception before the next attempt. Right answers route to the next prerequisite-clean step. Both feed the cognition profile.
Avoid-repeat ledger
A 30-day ledger of every question a child has seen. The composer prefers unseen items so practice stays fresh — even across multiple sessions on the same skill.
Why verifier-first matters for KS2
Adaptive engines that don't verify their output are guessing twice — first whether the question is right for the child, then whether the question itself is right. KS2 questions look simple to adults but a single wrong answer keyed against a child's previous mistake reinforces the misconception, not the curriculum. Northstar's gate runs a closed rule-set per micro-skill (40+ across Maths, SPaG, and Reading). If the gate cannot prove the candidate is correct, the candidate is dropped and the deterministic engine fills the slot. The child never sees the difference; the platform never ships a wrong answer it can't justify.
What changes when learning is truly adaptive
Three things parents and teachers consistently notice in the first two weeks. First, sessions feel shorter — children spend less time on questions that are too easy or too hard. Second, misconceptions surface earlier — the weekly parent report names the specific skill (not just the topic) where the child is drifting. Third, the route to SATs gets clearer — by Year 6 the cognition profile aggregates two or three years of evidence, so the SATs preparation plan is built from real data, not a generic revision list.
Adaptive practice across KS2 — year by year
Adaptation looks different at each year because the curriculum looks different.
Year 2
Calibration questions are simple — does the child have place value secure? — and adaptation focuses on the four operations + early measurement.
Year 3
Drift detection on times tables and column methods. The engine starts using cross-topic prerequisites (e.g. fractions needs division).
Year 4
Misconception families per topic — comparing fractions, time word problems, sentence types. Reports identify the family, not just the topic.
Year 5
Prerequisite chains become deeper — percentages depend on fractions which depend on division. Mastery loops route across topics.
Year 6
Full SATs-ready preparation. The cognition profile carries two-plus years of evidence into the revision plan.
Frequently asked questions
What's the difference between adaptive learning and personalised learning?
Personalised learning is broader — pace, content, format. Adaptive learning is a subset: the system uses real-time signals to choose the next question. Northstar is both. The curriculum journey is personalised by year and topic; the question-level decisions are adaptive based on the cognition profile.
Does Northstar use AI?
Yes — but every AI-generated question is re-verified by a closed-rule gate before it reaches a child. The gate is the source of truth for correctness; the AI is only ever a candidate generator. If the gate can't verify the candidate, the deterministic engine ships the slot.
Is Northstar suitable for SEND learners?
The adaptive engine reduces frustration by avoiding questions that are too hard or too easy. Teachers can also pin a child to a specific year band so practice stays comfortable. We don't market Northstar as a SEND-specific tool — it's a mainstream KS2 platform that adapts.
How does Northstar decide a child has mastered a skill?
Three signals: a minimum number of correct responses at the right cognitive demand level, recency (mastery decays without practice — the avoid-repeat ledger spots when something hasn't been seen for a while), and no fresh drift in the most recent session. The cognition profile holds all three.
Is Northstar curriculum-aligned?
Yes — every question maps to the UK National Curriculum at the year, topic, and skill level. The /learn surface mirrors that taxonomy directly, so teachers can see exactly which curriculum cells are covered.
Can I see what my child is being asked?
Yes — parent reports include sample questions per topic + the explanation each child saw. Teachers see the same view at the class level.