Learning Sciences
If you could give your child only one of two gifts, which would you choose?
The first is knowledge: a lifetime's worth of it, accurate, well organised, ready to use. The second is the ability to find their way through knowledge coherently — to tell what still holds from what has gone stale, what matters from what is merely well supported, and where they stand when the ground keeps moving.
If the choice felt easy, that is the gap this page is about. Education is built, almost entirely, to deliver the first gift. It measures it, ranks it, and certifies it. The second gift is the one the system has no instrument for.
The Paradigm We Inherited
For most of human history, one lifetime of learning was mostly enough to keep up. A skill learned young could still be trusted at fifty. Under those conditions, acquiring knowledge and being oriented in the world were close enough to the same thing that nobody needed to tell them apart.
They are no longer the same thing. The informational environment can now change faster than a single mind, unaided, can track it — the same rate mismatch that once made genetic adaptation too slow for a changing world, now appearing inside a single lifetime.
→ It Takes All the Running You Can Do
Facts Go Stale Without Telling Anyone
A fact does not announce the day it stops being true. A conclusion that was sound when it was learned goes on looking sound long after the conditions that justified it have changed, which is how a well-stocked head can be badly oriented. Having more information is not the same as knowing where you are.
The old saying about fishing already contains this. A fish is knowledge handed over complete: real, useful, and gone by tomorrow. Fishing is learning to read the river as it is today, because the river never held still long enough to be learned once. A school that hands out maps, however good, without teaching the checking of a map against the ground is handing out last year's river.
→ The River You Step In Twice
→ The Map Is Not the Territory
→ The Problem, where the difference between information and orientation is laid out in full
What Generative AI Changed
Generative AI did not create this problem. It made it impossible to ignore. When a student can produce a fluent essay, a worked solution, or a confident analysis in seconds, the finished product no longer shows whether anyone was oriented while producing it. Assessment has always leaned on the product as a proxy for the reasoning behind it. The proxy has stopped working.
The first responses — detection software, prohibition — aim at the product. Neither touches what has actually been lost: visibility of how the learner's thinking developed. What they asked, what they assumed, what they revised, and where their own judgment took over from a suggestion. This exact gap, and one experimental response to it, is worked out in full in Education After Generative AI.
A Hand on the Rudder
A learner using AI is a sailor with a far bigger engine. More power does not keep a boat on course. A hand on the rudder does, correcting continuously, in small adjustments, for as long as the voyage lasts. The learner's job — the one no tool can take over without taking the learning with it — is to remain the one who decides which way the reasoning goes.
That is what checkpoints in a learning process are for. Not surveillance: practice. The repeated moment when a learner stops, looks at what the tool produced, and decides what to adopt, what to reject, and why. It is also where human agency lives inside AI-assisted learning.
What a Learning Record Would Show
If orientation is the thing worth teaching, then a record of orientation is the thing worth keeping. Not only the final answer but the path to it: the questions that came first, the assumptions held and later dropped, the moments understanding changed, where the learner overrode the tool and where they didn't. A teacher reading that can see whether a student can find their way, not only whether they arrived.
→ Pay-It-Forward Record (PIFR)
It also means earlier work does not vanish when the course ends. Later learning can start from it instead of resetting: understanding that compounds because it is reused.
Where This Stands
The architecture behind this includes a deployment context built for educational use. What it does not yet have is the evidence that matters most to this audience: empirical studies, in real learning environments, of whether preserved reasoning actually improves how students learn. That is the question this research programme wants answered, not one it claims to have settled. The diagnosis above does not wait on it.
It will also take years. How education defines learning moves at the speed of culture, not software. A generation of teachers, assessors, and parents who were measured on acquisition now has to decide what it would rather measure. Nothing here argues for waiting. It argues for starting with the question at the top of this page.
Nobody needs convincing which gift they would choose. The work is building an education that gives it.
→ Why this matters to the author: Before It Had a Name