EdTech & the Science of Learning

01

Teaching as a system

The instinct when teaching is to make everything as smooth and easy as possible — clear slides, worked examples, nothing confusing. That instinct is half right and half disastrous. Some friction helps learning and some hurts it, and the whole art is telling them apart. The science sorts cleanly into two buckets:

• Reduce the friction that wastes effort — confusing layout, too much at once, split attention. This is cognitive load.
• Keep the friction that builds memory — effortful recall, spacing, mixing topics. These are desirable difficulties.

Everything below is one or the other.

02

Cognitive load: the bottleneck

Working memory — the mental space where you actively think — is tiny. It holds only a few items at once and empties in seconds. Cognitive Load Theory says all learning is bottlenecked there, and splits the load into three kinds:

• Intrinsic — the inherent difficulty of the material (gradient descent is just harder than a bar chart). You can't remove it, but you can sequence it.
• Extraneous — load from how it's presented: a cluttered slide, a diagram whose label is on the next page, jargon used before it's defined. This is pure waste, and cutting it is the single biggest lever a teacher has.
• Germane — the good load: the effort of actually building understanding. This is what you want learners spending their scarce capacity on.

The practical moves fall straight out of this:

• Chunk. Break material into small pieces and build up. Don't show the whole architecture at once; reveal it a layer at a time.
• Worked examples first. For novices, a fully worked solution teaches more than struggling with a blank problem — it shows the path before asking them to walk it.
• Kill split attention. Put the label on the diagram, not in a legend elsewhere; narrate a visual rather than making people read and look at once.

03

Dual coding: words and pictures

Dual coding theory says we process verbal and visual information through two separate channels, so a clear diagram paired with a clear explanation gives the brain two complementary routes to the same idea — and roughly doubles the working-memory budget instead of overloading one channel. It's why every page in this section pairs an SVG with prose rather than relying on either alone.

04

Desirable difficulties: why easy fails

Here's the most counter-intuitive finding in the whole field, from Robert Bjork: conditions that make learning feel harder and slower often make it stronger and more lasting. Re-reading notes feels productive — it's smooth, familiar, you recognise everything — but recognition isn't memory, and that fluency is an illusion. The techniques that actually build durable knowledge feel like more effort precisely because they are, and that effort is the mechanism.

05

Spacing & retrieval: the two big ones

Two desirable difficulties carry most of the weight, and they compound:

• Retrieval practice (the testing effect) — the act of pulling information out of memory strengthens it far more than putting it in again. A low-stakes quiz, a flashcard, or just closing the book and writing what you remember beats re-reading by a wide margin. Every "refresh in 60 seconds" box in this section is a deliberate retrieval cue, not a summary.
• Spacing (distributed practice) — the same study time spread across days beats one cram session. Each time you let memory fade a little and then retrieve it, it comes back stronger; this is the basis of the spacing curve.

Put together they become spaced retrieval — revisiting material at expanding intervals, recalling it each time — which is the single most evidence-backed study method there is, and exactly what spaced-repetition apps automate.

06

Interleaving: mix the problems

The instinct is to drill one skill to mastery (all gradient-descent problems, then all regularisation problems) — blocked practice. Interleaving mixes them instead, and reliably wins for anything where you later have to choose the right method. Blocked practice lets you run on autopilot — you already know every problem on this page is the same type. Mixed practice forces you to first ask "what kind of problem is this?", which is exactly the discrimination skill real work demands. It feels worse and scores lower in practice, then transfers far better — a desirable difficulty through and through.

07

What EdTech actually adds

Technology doesn't replace these principles — at its best it operationalises them at a scale a human teacher can't:

• Spaced-repetition systems (Anki and the like) schedule retrieval at the optimal moment per item, per learner — spacing + retrieval, automated.
• Adaptive learning adjusts difficulty to keep each learner in the productive zone — not so easy it's idle, not so hard it overloads — personalising intrinsic load.
• Immediate feedback closes the loop fast, so a misconception is caught before it sets.
• Learning analytics — the dashboards and data behind the platform — show where a cohort is struggling so teaching can adapt.

08

How I taught it

09

Refresh in 60 seconds

Principles reflect established learning-science research (cognitive load — Sweller; desirable difficulties — Bjork; the testing and spacing effects) alongside hands-on peer-mentoring and EdTech work.