ΣigmaClava

Programs of inquiry · term 03

Learning as Experience

AI-shaped curricula and learning environments that adapt to the child, not the child to them.

Active — field

The question

AI can already generate curriculum. The open problem is the experience of learning: environments that adapt to the child — pace, modality, interest, regulation — not the child to them.

What is at stake

Adaptive learning has mostly meant adaptive sequencing: the same experience, reordered. What a child actually meets is a texture — how it feels to be stuck, whether interest is treated as signal or noise, what happens at the moment attention goes. Generating more material does not touch any of that, and a system that mistakes coverage for learning will produce children who are well-covered.

A contour surface whose iso-lines crowd tightly around one point and, more faintly, a second — the surface deforming around a learner rather than the learner being flattened onto it.
Plate I The surface bends around the learner. Where the contours crowd, the environment is doing the adapting; the flat ground is where it has nothing to offer yet.

Why it is ours

Family-scale deployments give us a live field site for how children actually meet adaptive systems.

We are building AI-shaped learning environments and the instrumentation to understand what a child’s experience of them really is.

How we work on it

  • Instrument the experience, not the score.

    What we want to know is what the hour felt like — where interest caught, where it broke, what the child did when it broke — and scores are a poor proxy for any of it.

  • Interest is data.

    A child’s pull toward a subject is signal about how they learn, not a distraction from a syllabus. Systems that route around interest are discarding the most useful thing they have.

  • Regulation before instruction.

    A dysregulated child is not a child learning slowly; they are a child not learning. Environments have to notice this and change what they are doing, not push harder.

  • The child is not the variable.

    Where child and system disagree, the presumption is that the system is wrong. This is a design stance, and it is the expensive one.

What would count as an answer

A program is not finished when it produces results. It is finished when these are true, and can be shown to be true by someone who did not run it.

  • The system’s account of where a child struggled matches what an attentive adult in the room would say.
  • Interest that appears mid-session changes what happens next, within the session.
  • A child can be stuck without the environment escalating demand.
  • Removing the system does not remove the learning.

Illustrative studies

These describe the kind of study this program runs. They are illustrations of method, written to make the approach concrete — not work that has been completed, scheduled, or funded, and not findings. When something has actually been established, it appears in the Library with its method and its limits attached.

  1. 01

    The break point

    Design
    Sessions are reviewed around the moment attention breaks. What the environment did in the following minute is classified — escalate, hold, redirect, stop — and set against what happened in the next session.
    What it would read
    Whether the minute after attention breaks is the one that determines the hour. Our working assumption is that it is, and that most systems spend it badly.
  2. 02

    Interest routing

    Design
    When a child pulls toward an adjacent subject, the environment follows it rather than returning to the plan. Coverage and retention are compared against sessions that hold the plan.
    What it would read
    What following interest actually costs in coverage, and whether it buys enough in retention and willingness to return to be worth it.
  3. 03

    The attentive-adult standard

    Design
    An adult who was present writes what they observed. The system’s account of the same session is compared with it, blind.
    What it would read
    Whether instrumentation is measuring the experience or only the interaction. A system that disagrees with the adult in the room is making a claim it has to earn.

Terms this program uses

Learning environment
Everything a learner meets — pacing, texture, what happens when they are stuck, whether interest is treated as signal — as distinct from the material being covered. The environment is the part that determines whether the material lands.
Regulation-first
The stance that a dysregulated learner is not learning slowly but not learning, so the correct response is to change what the environment is doing rather than to press harder on the task.
Coverage trap
Mistaking material delivered for learning done. A system optimising for coverage will reliably produce a well-covered child, and the measurement it uses will agree with it.

The principles that bear hardest on this program are The human is the constant. and Experience is evidence. — see the principles and how intelligence is governed here.