👋 Hi {{first name | reader!}}
In this edition of ⚗️DistillED, we're looking at the Generation Effect — one of the most robust findings in memory research, and the idea that explains why AI is quietly undermining learning in ways that are easy to miss.
In this edition
📊 Do Now: Quick Poll!
When students use AI in your class, what are they most commonly doing with it?
What is the Generation Effect?
When students generate an answer — retrieve a word, complete a sentence, solve a problem — they remember it significantly better than when they simply read the same information presented to them. This is the Generation Effect: the act of production is itself the act of learning.
The phenomenon was named and systematically demonstrated by Slamecka and Graf in 1978. Across five experiments, they found that words participants generated from a cue — completing a word pair like RAPID: f___ — were consistently better remembered than words they simply read. The finding was striking in its simplicity: the act of producing the answer, not the correctness of it, was what drove the memory advantage.
In typical ⚗️DistillED fashion, here's what the Generation Effect is — and what it isn't:
👎 The Generation Effect IS NOT:
About making things harder for the sake of it → The struggle has to be the right kind — productive, relevant, and aimed at the right content
An argument against explanation → Teachers should still explain clearly. The sequence matters: attempt first, explanation after
A reason to avoid AI entirely → The issue is not AI itself. It is what AI is being asked to do
The same as trial and error → Generation works because it forces retrieval and construction, not because failure is inherently valuable
👍 The Generation Effect IS:
A memory phenomenon → Self-generated information is encoded more deeply than received information
Supported by decades of research → Replicated across 86 studies with a consistent effect
Immediately relevant to AI use → The way most students use AI bypasses generation entirely
Applicable across every subject → Wherever students are asked to think, the generation effect applies
Paul Kirschner captures the distinction that makes this matter right now:
"What people are doing with AI is something fundamentally different. They aren't storing intermediate results so they can continue thinking. They're handing over the thinking itself. That's cognitive outsourcing."
The only question is whether the tasks you set protect generation — or inadvertently remove it.
Why Does Generation Deepen Learning?
The problem is that receiving information feels like learning. A student reads a clear explanation and thinks they understand it. They watch a fluent video and feel like they've got it. They ask an AI and read the answer back. None of these feel like failure — but none of them trigger the cognitive work that makes knowledge stick.
A meta-analysis of 86 studies confirmed that self-generated information was remembered significantly better than information that was simply read (Bertsch et al., 2007). The struggle of production is not an obstacle to learning. It is the learning.
This is where Kirschner's distinction between cognitive offloading and cognitive outsourcing becomes essential.
Offloading — using a diagram, writing notes, working through steps on paper — frees up working memory while keeping the student as the thinker. The artefact stores information so the mind can continue processing. You're still doing the thinking.
Outsourcing — asking AI to explain, summarise, write, or produce — transfers the thinking itself to an external agent. The student receives a finished product. Nothing is encoded in long-term memory.
The consequences are clearest in writing. When a student drafts an argument, they are not just producing text — they are retrieving, organising, and stress-testing what they know. The product is evidence of thinking that already happened. As Hendrick and James put it:
"Writing is thinking. It is the act of retrieving knowledge, wrestling with syntax, and organising logic that forges understanding. When AI produces the final text, the student possesses the answer but lacks the understanding of how it was derived."
The note-taking literature makes the mechanism concrete. Pre-AI research showed that constructing notes substantially improves comprehension — not because of the notes themselves, but because building them forces generative processing. AI note-generators remove that activity entirely. Students receive a summary they did not make.
👉 If you want the broader picture of how memory works and why effort matters, the edition on The Simple Model of the Mind is the place to start.
How Do I Apply This in My Classroom?
The HOW here is not about banning AI or adding more to your plate. It is about sequencing. Generation needs to come before delivery, and production needs to be protected from shortcuts — not because effort is virtuous in itself, but because the effort is the mechanism. These five moves are ways of keeping that mechanism intact.
Step | Explanation | Example |
|---|---|---|
1. Protect Generation Before Delivery
| Slamecka & Graf showed that the act of generating — even incorrectly — produces stronger memory than reading a correct answer. The struggle is not a precursor to learning. It is the learning. | Before teaching photosynthesis, ask: “How do you think plants make food if they cannot eat?” Students sketch or explain their thinking first. Only then is the formal explanation introduced. 👉 Link to Daily Review — the daily review routine is one of the cleanest implementations of generation before delivery |
2. Use Retrieval Before Re-exposure
| Bertsch et al’s meta-analysis of 86 studies confirmed that self-generated information is remembered significantly better than read information. The implication is not about handout format — it is about sequence. Before revisiting material, ask students to retrieve it first. | Before handing back a practice essay, students first write everything they can remember about the success criteria and topic from memory. Only afterwards do they reread notes, teacher feedback, or model responses. |
3. Distinguish Offloading from Outsourcing
| Kirschner draws a hard line: offloading — notes, diagrams, working on paper — frees working memory while keeping the student as the thinker. Outsourcing — asking AI to explain or produce — transfers the thinking entirely. Design tasks accordingly. | Offloading: students use a planning scaffold, formula sheet, or mind map while solving a problem themselves. Outsourcing: students ask AI to “write my explanation of the water cycle” and submit ideas they never generated. |
4. Make Note-Taking a Generative Act
| Hendrick argues that deep note-taking improves comprehension because it forces students to reformulate material in their own words — the construction of notes is itself the learning event. AI-generated summaries remove that activity entirely, leaving students with a product they did not build. | After a lesson, students close their books and reconstruct the main ideas from memory in their own words before checking against their notes. The reconstruction attempt is the learning event. |
5. Design Tasks Where the Process Cannot be Outsourced
| Hendrick & James argue that writing is thinking — the act of retrieving, wrestling with syntax, and organising logic forges understanding. When AI produces the text, the student possesses the answer but not the understanding. Protect tasks where the making is the point. | Instead of “Write a paragraph on volcanoes,” first ask students to explain volcanic eruptions live to a partner using only a blank diagram. The explanation process forces retrieval, organisation, and generation in real time. |
The Generation Effect is a description of how memory works. Students who generate are students who learn. The question worth sitting with this week is not whether your students are working — it is whether the work they are doing is theirs.
Until next week, — Jamie
If you want more:
👉 Read The Simple Model of the Mind — how memory works and why the effort of learning is not incidental to it
👉 Read Explaining With Precision — designing explanations that land, including how sequencing affects encoding
👉 Read Retrieval Practice — the single most evidence-backed strategy for strengthening memory over time
👉 Get the DistillED Playbooks for practical, evidence-informed classroom practices
📥 Free One-Page Guide
The Generation Effect One-Page Guide
This edition comes with a free one-page guide on the Generation Effect — a practical summary of what it is, why it matters, and how to protect it in your classroom. Print it, share it with your department, or use it as a planning reference before your next lesson.
⚗️DistillED+ Exclusive Content
This week, ⚗️DistillED+ members get access to the Generation Effect CPD PowerPoint and Strategy Map — breaking down the offloading vs. outsourcing distinction, how to audit your current tasks for generation, and a practical planning framework for protecting cognitive work in an AI classroom.
