II itself is not proven effective — that claim is not made anywhere on this site. What follows is published, independently checkable evidence behind specific teaching and learning methods, and this prototype's own design choices that apply them. Every method below carries an evidence class. Classes C, D and P are never presented as causal proof.
Concept only. No accounts, no real child data, no tracking. Sources are named as text with a verification date, not linked out — this static preview makes zero outbound calls.
Evidence classes
Five classes, in order of strength.
AMeta-analysis, systematic review, or randomised controlled trial (RCT).
BBroader evidence synthesis — several studies pointing the same way, not yet a formal meta-analysis.
CObservational association — real data, but correlation, not a controlled experiment.
DOfficial high-performing-system practice — what a real education ministry does, not a controlled comparison.
PII hypothesis to test — our own idea, not yet evidence at all.
C, D and P describe association, practice or intention — never proof of causation. Only A and, more cautiously, B support a causal "this method works" claim.
Methods used in the Learning Engine
Click a badge for the exact source.
Retrieval practice — recalling an answer from memory, rather than re-reading it, is one of the most consistently supported study techniques in the evidence.
Agarwal, P. K., Nunes, L. D., & Blunt, J. R. (2021). "Retrieval Practice Consistently Benefits Student Learning: A Systematic Review of Applied Research in Schools and Classrooms." Educational Psychology Review. Systematic review coding 50 experiments (49 effect sizes, total n = 5,374); 57% showed medium-or-large benefits from retrieval practice, across education levels and subjects. Verified 2026-09-16.
Spaced review — returning to a topic after a delay, rather than only massing practice in one sitting.
Carpenter, S. K., Pan, S. C., & Butler, A. C. (2022). "The Science of Effective Learning With Spacing and Retrieval Practice." Nature Reviews Psychology, 1(9), 496–511. DOI 10.1038/s44159-022-00089-1. Verified 2026-09-16.
Worked examples — seeing a complete solved example before attempting one alone, in mathematics specifically.
Barbieri, C. A., Miller-Cotto, D., Clerjuste, S. N., et al. (2023). "A Meta-analysis of the Worked Examples Effect on Mathematics Performance." Educational Psychology Review, 35. 43 articles / 55 studies / 181 effect sizes; medium average effect, g = 0.48 (p = 0.01). Correct-only worked examples outperformed examples mixing correct and incorrect. Verified 2026-09-16.
Interleaved practice — mixing different problem types instead of blocking one type at a time, in mathematics.
Rohrer, D., Dedrick, R. F., Hartwig, M. K., & Cheung, C.-N. (2020). "A Randomized Controlled Trial of Interleaved Mathematics Practice." Journal of Educational Psychology, 112(1), 40–52. 787 students, 54 classes, 5 schools; interleaved practice outscored blocked practice on a later test, 61% vs. 37% (Cohen's d = 0.83). Verified 2026-09-16.
Metacognition & self-regulation — explicitly teaching planning, monitoring and evaluating one's own learning.
Education Endowment Foundation (EEF), Teaching & Learning Toolkit — "Metacognition and self-regulation." Meta-analysis basis (355+ studies, updated with 107 more); average impact approximately +8 months' additional progress over a year, across subjects and age phases. Verified 2026-09-16.
Feedback — information about the task and strategy, not the person, given close to the attempt.
EEF, Teaching & Learning Toolkit — "Feedback." Average impact approximately +6 months' progress; evidence security rated "high." Feedback about the task and strategy outperforms feedback about the learner as a person. Verified 2026-09-16.
Mastery learning — fixed learning outcomes, variable time, before moving on.
EEF, Teaching & Learning Toolkit — "Mastery learning." Toolkit strand with a positive average impact; this prototype does not quote a specific months-of-progress figure for this strand because none was confirmed during verification — see the Toolkit itself for the current number. Verified 2026-09-16.
Small-group tuition — one teacher or tutor with 2–5 pupils together.
EEF, Teaching & Learning Toolkit — "Small group tuition." Average impact approximately +4 months' progress over a year. Verified 2026-09-16.
One-to-one tuition — short, regular individual sessions over a bounded period.
EEF, Teaching & Learning Toolkit — "One to one tuition." Average impact approximately +5 months' progress; effect size varies by subject (around +2 months in mathematics, +6 months in literacy in the underlying studies). Best results from short (~30 min), regular sessions over a set period (up to ~10 weeks). Verified 2026-09-16.
Reading comprehension strategies — explicit strategies for understanding a text.
EEF, Teaching & Learning Toolkit — "Reading comprehension strategies." Average impact approximately +7 months' progress. Verified 2026-09-16.
Oral language — structured spoken-language and discussion activities.
EEF, Teaching & Learning Toolkit — "Oral language interventions." Average impact approximately +5 months' progress, with positive knock-on effects for reading comprehension. Verified 2026-09-16.
Cross-national evidence — associational, not causal
PISA 2022 / OECD
OECD, PISA 2022 Results (Volume V): Learning Strategies and Attitudes for Life, published November 2024 (from the 2022 assessment cycle). A large cross-national survey — correlational, not an experiment. Verified 2026-09-16.
PISA evidence is associational, not causal — it shows what goes together across countries, not what causes what. Two concrete findings: fewer than half of OECD students report trying to relate new material to what they already learned; on average only 47% frequently seek clarification when confused in mathematics. Students with higher intrinsic motivation and self-efficacy are more likely to use effective learning strategies — an association, not proof that motivation alone produces strategy use.
Official system examples
What high-performing systems do
Official education-ministry bodies, named here for what they are, not for a specific unverified policy claim: Ministry of Education (Singapore); Ministry of Education and Research (Estonia); Finnish National Agency for Education / Opetushallitus (EDUFI), under Finland's Ministry of Education and Culture; Ministry of Education, Culture, Sports, Science and Technology (MEXT, Japan). Verified 2026-09-16.
That a ministry does something is a signal, not a controlled experiment — a high-performing system doing X is not proof X caused the performance. Shown as Class D on purpose: worth knowing, not worth treating as settled science.
A concept, not a label for children
Learn like a strong learner.
Not "smart children do this." Strong learning habits include: checking for mistakes, asking when unsure, connecting knowledge, considering multiple approaches, and managing your own progress — the same self-regulation skills in the Learning Engine, restated as things any learner can practise. The PISA evidence behind this framing is, again, associational, not causal: students who report these habits tend to do better; that does not prove the habit alone produced the result.
Pilot design
What a real pilot would measure.
BaselineImmediate learningDelayed retentionError typesTime / attempts to masteryAbility to explainSelf-monitoringConfidence / anxietyNeed for human support
Primary product metric: learning retained — not time on site. A design that maximises time spent would be optimising for the wrong thing.
Pilot phases
This is Phase 0. Nothing later is running yet.
Phase 0 — now
Static public concept. No child accounts or data of any kind.
Phase 1 — future
A small, explicitly consented pilot, gated on parent consent, legal review and privacy design that does not exist yet.
Phase 2 — future
A controlled comparison of learning methods. Neither Phase 1 nor Phase 2 is implied to be operating by anything on this site.