What Is High-Agency Learning?
High-agency learning gives young people meaningful choices, real responsibility, and repeated chances to turn their ideas into useful work. The learner chooses, frames, builds, tests, explains, and improves. Capability grows through those decisions.
- The learner makes the important decisions.
- AI scaffolds thinking. It does not replace it.
- Confidence comes from evidence, not praise.
1. Definition
High-agency learning gives young people meaningful choices, real responsibility, and repeated opportunities to turn ideas into useful work. They make consequential decisions, build, test, explain, reflect, and improve.
In most schooling, adults hold the interesting decisions: the goal, the method, and the standard of success. High-agency learning gives the learner as many of those decisions as they can handle, with enough support to succeed.
“High-agency learning is what happens when a young person gets to make real decisions, do the hard part themselves, and build something that matters to them — with just enough support to succeed.”
2. Why agency matters in an AI-rich world
AI can produce a plausible answer to almost any assigned task in seconds. So a finished task no longer proves the learner became more capable.
Agency matters more now: deciding what is worth doing, judging whether an answer is good, and continuing after a failed attempt. These skills grow through practice with real decisions. Fully automated help can remove that practice.
3. What high-agency learning is
In practice, a high-agency learning experience tends to include:
- Meaningful choice. The learner has real say over what they work on and how.
- Real responsibility. Decisions have consequences the learner can see.
- Something built. The work produces an artifact — an app, a model, a piece of writing — not just a grade.
- Room to struggle. Productive difficulty is left intact, with support nearby.
- Reflection and iteration. The learner explains their decisions and improves through feedback.
- Evidence of growth. Progress shows up as a record of what changed, not a one-time score.
4. What it is not
It is not “leave the kid alone.” Most learners need scaffolding to make good decisions. It is not just doing projects: a project with every choice pre-made is still low-agency. And it is not anti-AI. The goal is to keep the thinking with the learner.
A quick comparison
| Lower agency | Higher agency |
|---|---|
| An assigned goal | A meaningful learner choice |
| Completing a task | Solving a real problem |
| AI producing the answer | AI scaffolding the process |
| Praise-based confidence | Evidence-based confidence |
| A one-time output | A record of iteration and growth |
| Following instructions | Defending decisions |
5. The role of interest
Interest is the fuel. A young person who cares about a game, a sport, or a local problem accepts far more difficulty. Interest also makes a choice meaningful.
So high-agency learning starts from what the learner brings, then stretches toward skills they would not choose on their own.
6. The role of projects
Projects give agency somewhere to go. A real project forces decisions about scope, approach, trade-offs, and when it is done. The result makes growth visible.
The test: how many of the interesting decisions did the learner make?
7. The role of AI
My guiding principle is simple: AI should scaffold thinking, not surrender it. Used well, AI is a patient tutor and a way to get unstuck. Used badly, it hands out answers and removes the decisions where learning happens.
Good AI support asks questions, offers options, and hands the decision back. To test this, use the Scaffold or Surrender? framework.
8. The role of feedback and reflection
Skill grows when the learner tests the work, notices what failed, explains their choices, and decides what to change. Feedback turns one attempt into a loop.
A learner who can defend their reasoning has understanding that transfers. A learner who only produced an output may not.
9. Agency, competency, and confidence
These three grow together. Agency is the practice of making real decisions. Competency is the ability those decisions build. Earned confidence comes from evidence that the learner can handle the next hard thing.
Praise produces fragile confidence that breaks at the first setback. Evidence produces durable confidence, because it rests on what the learner actually did.
10. What this means for parents
Protect the struggle. Do not finish the hard part for your child, and do not let an AI do it. Ask what they are trying to do and what they tried. Let small failures stand. Praise the evidence: “You figured that out” lands differently than “good job.”
11. What this means for educators
Give students the interesting decisions where the stakes allow, and scaffold deliberately. Assess reasoning and revision, not only the final product. That is much harder to shortcut with a tool. Treat AI like a calculator: allowed, named, and bounded.
12. What this means for AI product builders
Your product's default behavior is a teaching choice. Instant answers optimize for task completion. Questions, options, and handing the decision back optimize for capability. Design for capability. A smoother experience is not the same as more learning.
13. How SayMake applies the ideas
SayMake is my AI-supported maker studio for young builders. It is an experiment. It starts from a kid's interest and guides a full build. The AI helps a kid get unstuck but leaves the important decisions with the kid.
The aim is a record of real projects a kid can point to. Its status is in the project archive (availability unverified). The build loop itself is documented in the Build to Become framework.
14. Sources and further reading
High-agency learning draws on established traditions. They overlap but are not interchangeable.
Learner agency
The capacity to make consequential choices and act on them. The umbrella idea.
Self-directed learning
The learner sets goals and manages progress. High-agency learning shares the focus on choice but welcomes scaffolding.
Project-based learning
Learning organized around building over time. A project can still be low-agency if every decision is made for the learner.
Competency-based learning
Progress measured by demonstrated ability, not time. High-agency learning also values the decisions that produce competency.
Constructionism
Seymour Papert's idea that people learn well by making shareable things. A close ancestor; high-agency learning adds choice, reflection, and AI's role.
Interest-driven learning
Learning that starts from what the learner cares about. Interest is the fuel.
Primary sources worth reading. They inform these ideas, not every specific claim.
- Seymour Papert, Mindstorms: Children, Computers, and Powerful Ideas (1980) — the foundational text on constructionism and learning by making.
- Edward L. Deci & Richard M. Ryan, self-determination theory — research on intrinsic motivation and the role of autonomy, competence, and relatedness.
- Brigid Barron, research on interest-driven and self-sustaining learning across the settings of a young person's life (learning ecologies).
Consult these works for their precise claims.