Skip to main content
lab-notes.aiPublic-service AI playbooks
Opportunity 01/17EducationEducation and learning

Adaptive Tutoring

Explore what a practice tool would need to record before choosing a learner's next activity, using topic-level data rather than learner records.

Inspect starter dataSet up the build partner
Published sources
2
Starter data
18 synthetic records
Domain build partner
build-adaptive-tutoring
Last reviewed

From the draft strategy

Opportunity

The draft strategy names adaptive material, tutoring, and interactive learning as a potential public-service application, suggesting it could suit different ways of learning, without saying what evidence the choice of next activity would rest on.

Where the draft says it

Table 2 — potential public-service applications

Read the draft strategy source (opens in a new tab)

Source register

Research already done

Real published source

We reviewed these published sources so you do not have to start from zero. Each source shows what it covers, how it can be accessed, and why it matters.

2 sources

  1. Department of Education

    Open

    School leavers

    What it covers
    Annual figures on the highest qualifications Northern Ireland school leavers achieve and where they go next, broken down by school type and by pupil characteristics.
    Why it matters here
    It is the published record of how attainment is spread here, which is the only evidence available about outcomes without going near an individual learner.

    Open the source (opens in a new tab)

  2. Department of Education

    Open

    School performance

    What it covers
    Annual examination performance bulletins for pupils in Year 12 and Year 14 across Northern Ireland post-primary schools.
    Why it matters here
    It shows what is published about how pupils do, and by contrast that nothing published describes how anyone practises, which is what this example would depend on.

    Open the source (opens in a new tab)

Starter data

Starter dataset

Synthetic working data

Use 18 synthetic topic-level practice summaries to explore how a tool might choose a next activity without learner records.

AI-assisted research helped identify and interpret the published sources. We then created a small, non-sensitive synthetic dataset shaped by the information those sources expose. It is for exploration—not evidence, training, or operational decisions.

Records
18
Fields
5

Field preview

table view
  • id

    SYN-AT-01 · SYN-AT-02 · SYN-AT-03

  • topic

    Equivalent fractions · Fractions of amounts · Rounding to significant figures

  • difficulty

    foundation · developing · secure

  • attemptsBand

    500-1000 · 250-500 · 100-250

  • successRateBand

    70-85% · 55-70% · over 85%

Inspect all 18 recordsView repository file (opens in a new tab)Download JSON (opens in a new tab)

How it was prepared

AI authored fictional topic and difficulty steps with banded attempt counts and success rates.

Limitations

  • The bands were invented for this project and are not measurements, so no figure here is evidence about how hard any topic really is.
  • Everything is aggregated by topic, which is the safe choice and also removes the sequence of attempts that a real adaptive tool would work from.
  • Eighteen topics from one subject area cannot stand in for a curriculum, and there is no reviewed bank of activities behind them.

Clone and build

Domain build partner

Build with a domain-aware coding agent

Clone the repository, then ask your coding agent to load the skill that ships with it. The skill brings the opportunity, sources, starter data, known unknowns and constraints into the conversation before anything is proposed. It follows the open Agent Skills standard, so it works in Claude Code, Codex and any other agent that reads skills.

01 / Clone

git clone https://github.com/Hypership-Software/ats-us-nai.git

02 / Invoke

Claude Code
/build-adaptive-tutoring
Codex
$build-adaptive-tutoring

Conditions, not a checklist

Before you build

These are not footnotes. They are conditions that any responsible prototype would need to address.

  • Learner data reveals sensitive details

    Practice data about children can reveal age, ability, additional needs, and behaviour, which is why this file counts attempts by topic and never by learner.

  • Small choices can narrow opportunity

    A tool that keeps choosing the wrong next step can narrow what a learner is ever offered, and each individual choice looks too small to notice.

  • Practice scores do not prove understanding

    Attempts and success rates say nothing about whether anyone understood the topic, and a low score may only mean the wording, the language, or the device got in the way.

  • Learning styles are not a sound basis

    Sorting learners by a preferred way of learning is not a sound basis for deciding what they see next, whatever the tool reports.