Track assistance and test a small amount of transfer. Do not try to infer authorship or competence from the polish of a finished project.
Use a contribution record, not surveillance
After each session, complete a few lines:
| Component or decision | Learner contribution | Mentor contribution | AI contribution | What the learner can now explain or change |
|---|---|---|---|---|
| Example: data-loading function | Chose fields and changed a filter | Explained an error | Suggested an initial code draft | Can alter the filter and explain the result |
The purpose is honest attribution, not maximising a percentage labelled “child-written.” A learner can legitimately use assistance while still developing understanding; the record should make that assistance visible.
What should a useful parent report contain?
Ask for five short items:
- Goal practised: the capability, not just the feature completed.
- Learner action: something the child actually explained, decided or changed.
- Support used: demonstration, hint, code, AI draft or direct intervention.
- Current difficulty: one specific unresolved issue.
- Next step: the next small task and any required preparation.
A report saying only “excellent engagement; chatbot completed” is insufficient for this buying decision.
Try the 15-minute “explain, change, debug” check
Use familiar material and tell the learner the purpose beforehand:
- Explain: “What information goes in, what comes out, and where might it fail?”
- Change: “Can you make this small adjustment?”
- Debug: “What would you inspect first if this result looked wrong?”
- Reflect: “What did the mentor or AI supply that you could not yet do alone?”
Record the support needed: independent, verbal hint, worked example or direct completion. Do not use the result as a diagnosis or compare children against each other.
About a week later, invite a tiny related task with a changed input. This is a proposed household check of retention and transfer, not a validated test. Keep it brief and supportive.
For parents wanting background on this distinction, the January 2026 Screen Deep episode with learning-sciences researcher Adam Dubé discusses choosing educational technology and distinguishing learning support from offloading the work. Its page includes audio, a transcript and a YouTube route. (childrenandscreens.org)
Sources cited on this page
- Does AI Help Students Learn? | Adam Dubé, PhD - Children and Screens ↗childrenandscreens.org