The desk recommends a small, reversible trial with an explicit learning review. The four-session structure below is a suggested purchasing format, not a research-established optimal dosage. Adjust session length and pacing to the learner.
| Stage | Suggested activity | Evidence to retain |
|---|---|---|
| Before session 1 | Close all gates; confirm the named mentor, cost cap, scope and review date. | One-page approved trial brief. |
| Session 1: Establish the baseline | Discuss the learner’s interest and attempt a small relevant task. | Starting skills, support needed and learner feedback. |
| Session 2: Build one small component | Let the learner make an identifiable decision and implement or test it. | Brief contribution note distinguishing learner, mentor and AI assistance. |
| Session 3: Change and check | Introduce a small new requirement, unexpected example or manageable fault. | How the learner reasons, tests and responds to help. |
| Session 4: Review and decide | Repeat a comparable small task and discuss enjoyment, challenge and next steps. | Continue, reduce scope, change mentor or stop—with reasons. |
Use notes and appropriately redacted work samples rather than making recording compulsory. Do not turn the review into a surprise examination.
For a non-technical parent, three useful prompts are:
- Explain: “What is this part supposed to do, and where might it be wrong?”
- Change: “What would you change if we wanted a slightly different result?”
- Check: “How could you tell whether your change worked?”
Ask the mentor to help design a fair task, but not to answer it for the learner. Record the level of help rather than assigning a pass/fail label to the child.
Sources and context
This page sets out the desk’s purchasing framework. Sources →