The learning techniques that work don't need a screen

Which study techniques actually raise attainment?

Five do most of the work: spaced retrieval, self-testing, interleaving, explaining aloud, and teaching someone else.

The evidence here is not new and it is not contested. It has been replicated for decades, and the ranking barely moves between meta-analyses. What is striking is how little of it survives into the products people actually buy.

Why does this matter commercially?

Because the techniques with the weakest evidence are the ones the industry is built on.

Video lectures are cheap to produce, easy to demonstrate in a sales meeting, and trivially measurable as watch time. They are also, of the ten techniques ranked, among the weakest. Re-reading and highlighting sit beside them.

So a category built almost entirely on recorded video is a category built on the bottom of the evidence table. That is not a hot take; it is what a 242-study meta-analysis says when you line the techniques up.

The uncomfortable implication is that the display — the thing every product in the category is organised around — contributes almost nothing to the mechanisms that make learning stick. It is a delivery convenience that got mistaken for a pedagogy.

Has the market moved on screens?

Yes, decisively, and policy moved with it.

Screen-free has stopped being a wellness preference and become a purchase criterion. That is a meaningful distinction: a preference influences which product a parent likes, a criterion decides which products they will consider at all.

• 58% of countries — 114 school systems — now restrict phones in schools, up from 24% in mid-2023. South Korea legislated a national ban.
• India's Economic Survey 2025–26 described compulsive scrolling as a public-health concern and recommended tech-free zones and digital-wellness curricula.
• 61% of urban Indian parents surveyed say their child is addicted to a screen; 39% report increased aggression.
• Screen-free products are growing where screen-based ones are not: Yoto up 86% year on year, Skillmatics up 87% in India.
• After the collapse of the sector's largest video-first company, the major Indian players moved hybrid or offline, because parents pay for learning they can see.

Does paper actually beat a screen, or is that nostalgia?

For expository text — the kind used to learn from — the paper advantage is consistent across four meta-analyses from 2018 to 2024.

Comprehension is measurably higher on paper, and the gap is widest precisely for the dense, explanatory material that schoolwork consists of. Handwriting engages wider brain networks than typing, though the retention benefit specifically is still contested and should be claimed carefully.

None of this means screens are useless. It means the screen is doing something other than helping the learner absorb the material, and a product should be honest about which of the two it is optimising for.

What does the parent evidence say?

Parental homework involvement is slightly negative overall — about r = −0.06 — and content-correcting is worse than that.

Autonomy-supportive involvement helps. Controlling, correcting involvement hurts, at around r = −0.15 in mathematics. A parent with a red pen is, on average, making things worse.

This has a direct design consequence. The parent should be given a short coaching ritual and a question to ask, never a marking task. Let the system do the correcting and leave the parent the part that actually helps.

It also suggests the cheapest intervention in the field is underused: a randomised trial of daily WhatsApp nudges to parents produced a 0.2 SD cognitive gain in five months, at roughly a dollar per child.

Where is the gap in the market?

Every quadrant is occupied except one: adaptive AI that operates entirely through paper and voice.

Nobody is putting the AI behind the paper. Competitors either put the model on the child's screen, or run an offline programme with no model in it. The empty quadrant is a curriculum-bound, adaptive system for ages six to eighteen where the child's interface is a sheet and a voice, and the parent is a coach rather than a marker.

That is the quadrant we are building for. What we are building has no display anywhere on it, and that is as much as we will say before it ships.

Video platforms

One to three hours of screen a day, passive intake, and parents increasingly hostile to the medium itself.

Photo-solve apps

Enormous daily usage, no learning mechanism at all, and schools are actively blocking them.

Chat AI tutors

Sometimes genuinely Socratic, but undifferentiated from each other and still a screen.

Offline coaching centres

Effective and expensive, with no loop that continues at home.

Paper worksheet franchises

Paper, and durable, but a fixed ladder with no adaptivity, no voice and no parent coaching.

Screen-free play kits

Excellent for young children, but they stop around age eight to ten and carry no exam relevance.

AI grading for schools

Assessment only. It grades what already happened rather than generating what comes next.

What should you take from this if you are building in education?

Design for the mechanism, then choose the medium — most teams do it the other way round.

• Pick the technique first from the effect-size table, then ask what hardware it actually requires. Often the answer is very little.
• Treat watch time as a cost, not a metric. Ranking by learning value and completion is a different product, not a setting.
• If your product needs a screen, be able to say what the screen is contributing that paper could not.
• Give the parent a script, never a red pen.
• Be careful with the weaker claims. Rhythmic recitation shows structural brain differences in lifelong practitioners, but there are no transfer trials for schoolwork. Use it for tables and formulas, and market it honestly.