Comparisons

How do you tell whether the child is learning rather than producing AI-generated work?

Compare the learner’s capability before and after—not just the final artifact. The distinction is consistent with the research finding that supported performance and later independent performance can diverge. (pmc.ncbi.nlm.nih.gov)

Explore the comparison in this briefing

Point in the engagement

01 / 7 options

Before teaching

Evidence to request
A short explanation, coding sample or practical task
What it helps distinguish
Existing capability versus later development

Compare each option against the same criteria. The complete table and qualifications remain in the briefing.

Our proposed six-session evidence plan is below. It is an adjustable purchasing and teaching framework, not a guarantee that every project should finish in six meetings.

Point in the engagement Evidence to request What it helps distinguish
Before teaching A short explanation, coding sample or practical task Existing capability versus later development
Session 1 A project brief in the learner’s own words Personal understanding versus repeating the mentor’s plan
Session 2 One small component and its explanation Understanding a component versus running a supplied project
Session 3 A change to a requirement Flexible understanding versus memorized steps
Session 4 A documented failure and response Evaluation and debugging versus demonstrations that only show successes
Session 5 Contribution notes identifying learner, mentor and AI assistance Honest attribution versus ambiguous ownership
Session 6 A related task attempted with reduced help, plus a teach-back Transferable capability versus familiarity with one artifact

For a non-technical parent, use four questions:

  • What did you decide yourself?
  • What did the mentor or AI provide?
  • What goes wrong, and how do you know?
  • Can you show one small change and explain its effect?

Keep the conversation brief and supportive. Preserve useful notes and project versions rather than making continuous recording or surveillance the default.

Should you renew, reset or replace?

Use the following decision rule, agreed before the first block ends:

  • Renew: the learner remains interested, agreed capabilities are becoming visible and operating arrangements work.
  • Reset: rapport is good, but the project is too broad, foundations are missing or evidence is unclear. Narrow the scope and agree a short reassessment period.
  • Replace: the teaching relationship remains unsuitable, commitments repeatedly fail or the instructor will not explain contributions and progress.
  • Pause immediately: a safeguarding or account-eligibility concern remains unresolved. Do not finish a prepaid block merely to use the balance.

Sources cited on this page

  1. Generative AI without guardrails can harm learning: Evidence from high school mathematics - PMC ↗pmc.ncbi.nlm.nih.gov