Skip to content
The Briefing: five minutes on what moved, every Monday. Subscribe free Editorial standardsCorrectionsContact
The EdJournal Independent research, forecasts and tools on the future of school
Menu
QuestionsAIToolsResearchForecastsJournalThe BriefingAboutWork with us

Questions  /  05 · AI

Question 05 · A living dossier

Where should AI sit in a developing mind?

Updated 25 September 2026Version 1.0Stands on The Formed Mind (PP-001), The State of AI Use and What FollowsBy Shilpa Kapur SinghCite this

The current answer

AI should extend a developing mind, not replace the effort that forms it, and support a young person’s relationships, not stand in for them. So let it in where the capacity it touches is already formed, and keep the effort where the capacity is still being built.

Whether young people will use AI is settled; they already do. What a school still decides is when, and in what order. That is a question about how a child develops, not about technology.

How this dossier was built

01
The State of AI Use
The baseline: what is actually happening in classrooms right now, measured. Research report RR-AI-01 →
02
What Follows
The extrapolation: where those trend lines are heading, and the one question they cannot yet answer. Forecast FC-AI-01 →
03
The Formed Mind
The argument: what that trajectory commits a school to, and why timing is the question that matters. Position paper PP-001 →
04
This dossier
The living framework: the working answer schools can act on now, kept current as the evidence changes. You are reading it.

The evidence is a childhood away

One childhood, 2026 to 2041A child who starts school at four in 2026 is nineteen in 2041, when the evidence on growing up with AI would arrive.2026age 42031age 92036age 142041age 19We are hereStarts school, age 4The evidence arrivesLeaves school, age 19Every year of the bar is a year of the child’s schooling spent inside the question.

The evidence that would settle this is fifteen years away. A childhood is the unit of measurement, and we are inside one.

So we stage rather than wait or rush, because the two ways of being wrong are not equal.

If we wait, and AI proves harmless

Children lose some years with a useful tool. They can pick it up later. That cost is recoverable.

If we rush, and AI replaces the effort that builds a capacity

A capacity is not built in the years it should have been. For that child, the window does not come round again. That cost may not be recoverable.

Two pillars

Read The Formed Mind

01

Extend the mind. Do not replace the effort that forms it. A capacity still being built cannot safely be offloaded, because offloading it removes the effort that builds it.

The cognitive pillar

02

Support the person. Do not stand in for their relationships. The window for forming the social capacities of adolescence is adolescence itself. It cannot be deferred.

The relational pillar

~700M

weekly ChatGPT users by mid-2025, close to one in ten people on earth. Adoption at that speed means schools decide after use has begun, not before.

OpenAI, via The State of AI Use

What to do first

01
Map where AI meets a capacity still forming
List the tasks where students now use AI, and mark the ones that touch a skill they are still building: writing an argument, working a problem, recalling what they know.
02
Keep the effort in the task
Where the capacity is still forming, design the work so AI extends the student’s thinking after they have done it, not instead of it.
03
Protect the relationships
Decide where AI must not stand in for a trusted adult or a peer, and say so in policy, especially for companion-style tools.
04
Stage it, and review every year
Set out which uses open at which age, and revisit the stages each September as the evidence arrives. Staging is a decision you can change; an unmade decision is not.

The forecast behind this dossier · FC-AI-01

They will use it for judgment before judgment has formed.

Read the forecastThe State of AI Use

What would change our mind

  • Long-run evidence that heavy AI use in childhood leaves the capacities it touches as strong as they would otherwise have been.
  • Evidence that AI companions widen rather than narrow young people’s relationships with other people.
  • A staged approach producing worse outcomes than open access, measured over years, not weeks.

Where this comes from

What changed

  • 28 Sept 2026
    Added DOI for The Formed Mind (PP-001), version 2; added the story of how this dossier was built; added structured data for search and AI assistants.
  • 25 Sept 2026
    Published as a living dossier, version 1.0.
  • Aug 2026
    Forecast FC-AI-01, What Follows, published.
  • July 2026
    Position paper PP-001, The Formed Mind, published.
  • June 2026
    The State of AI Use published.

How to cite

Kapur Singh, S. (2026). Where should AI sit in a developing mind? Living dossier, version 1.0, 25 September 2026. The EdJournal.