Critical friendship, companionship and human judgement in the AI era
An old folder contained questions that turned out not to be old at all.
What happens when somebody — or something — becomes very good at helping us think?
When does help enlarge our judgement?
When does it begin to take the work over?
And if the work gets better, what happens to the person doing it?
This series began by accident.
A folder shared years ago by Keith Lyons reopened a trail through critical friendship: challenge and support, ownership and authority, trust and independence, the difficulty hidden inside the word friendship.
One citation led to Angie Titchen and critical companionship.
That changed the question.
A return to Keith’s later idea of the Travel Companion changed it again.
Then the trail reached a much newer relationship:
Human:AI.
The point is not that AI is a critical friend.
It is not that Keith anticipated ChatGPT.
And it is not that an old human relationship can simply be copied across the colon.
The question is more useful than that:
What can older traditions of helping another person learn teach us about preserving human judgement when AI becomes extraordinarily capable at helping?
The journey unfolds across eight movements.
The movements are fixed.
The number of steps inside them was earned by the inquiry.
Read from the beginning and the question changes as the evidence changes it.
An accident becomes an inquiry
The series begins before there is a theory.
A folder opens.
The resemblance to current Human:AI practice is tempting.
The first discipline is not to trust it.
Prelude — Serendipity Opens the Door
The starting point: an accidental return to Keith’s Critical Friend folder and a decision to understand what was actually there before following the apparent connections.
Question carried forward:
What had Keith actually been looking at?
Before AI. Before companionship. What was critical friendship trying to solve?
The first movement stays inside Keith’s collected literature.
It finds no single definition.
Instead, the problem gradually changes:
seeing → ownership → enactment → friendship
1 — Inside the Critical Friend Folder
Another person can help us see our work differently. But useful help should also develop the learner’s capacity to see for themselves.
2 — Whose Friend Are You?
Once ownership matters, authority enters. Whose purposes does the relationship serve? Who authorises the challenge? Who ultimately judges the work?
3 — Learning to Be Critical
A role can be agreed before anybody has learned how to practise it. Challenge is not a volume control. Criticality has to be learned together.
4 — What Makes This a Friendship?
Useful professional critique is not automatically friendship. The word itself becomes part of the problem.
Question carried forward:
What did companion mean in the neighbouring literature?
A neighbouring tradition
Gibbs and Angelides leave another word in view:
companion.
Their reference list points to Angie Titchen.
The trail follows the citation — without pretending Keith did.
5 — One Citation Away
Titchen’s work is encountered on its own terms. The resemblance is noticed, but lineage is refused.
6 — Why Companion?
Companion turns out to carry more than friendliness: commitment, mutuality, reciprocity, particularity and the difficult little word with.
7 — A Companion Has to Leave
Accompaniment is not permanent presence. Development, capability and mutual parting are built into the relationship.
Question carried forward:
What happens if critical companionship is placed beside Keith’s later Travel Companion without pretending that one produced the other?
Back to Keith — but with a different question
The journey returns to material already encountered in Beyond the Breadcrumbs.
It does not rewrite it.
It sees it differently.
8 — A Connection We Cannot Claim
The attractive Keith → Titchen origin story fails the evidence.
A route between documents is not a route through somebody’s thinking.
9 — Was Travel Companion Really a Relationship?
Keith’s phrase becomes meaningful only through what happened around it: return, absence, challenge, listening, confidentiality, correction, and the other person’s freedom not to lean.
10 — The Direction of Learning Changed
The experienced person was not outside the learning.
Contribution could travel both ways without making the relationship symmetrical.
Reciprocity without symmetry.
Question carried forward:
Was that relational shape already present in our Human:AI practice before we knew to look for it here?
Human:AI enters the inquiry — carefully
Before comparing the old traditions with AI, the chronology has to be frozen.
What was genuinely present before the folder reopened?
What arrived later?
And what are we now in danger of reading backwards?
11 — Before We Opened the Folder
The pre-discovery record shows a Human:AI practice already concerned with rollback, refusal, reality contact, contribution without authority, and stewardship of judgement.
But the provenance is mixed.
Some behaviour is older.
Some later language was already being shaped inside a Keith-informed project.
12 — A Family Resemblance Is Not a Family Tree
The resemblance is genuine.
That does not make it ancestry.
Human critical friendship, critical companionship and Human:AI can illuminate one another without becoming the same relationship.
13 — Companionship Without Taking the Chair
The local phrase is tested.
AI can contribute materially to inquiry.
The Human retains consequential judgement.
Contribution is not authority.
Question carried forward:
What happened to critical friendship after Keith’s folder ended?
The literature did not stop in 2016
Before AI is allowed into the critical-friend role, the later literature makes the role itself harder to claim.
The question changes again:
definition → configuration → inspectable contribution
14 — The Question Did Not Go Away
Later researchers discover that critical friend has become too variable to explain itself.
15 — A Continuum, Not a Type
The relationship is better understood as a configuration than a fixed category.
Who is involved? How close are they? What do they know? What are they expected to do?
16 — Naming It Is Not Enough
Calling somebody a critical friend is not evidence that critical friendship happened.
The contribution has to leave a trace.
Question carried forward:
What happens when the critical friend becomes AI?
The old question reaches the AI era
Only now does AI enter directly.
And when it does, the question fractures.
AI can perform some of the work historically associated with critical friendship.
That does not settle what kind of relationship has been created.
17 — When the Critical Friend Became AI
AI is tested directly against an established critical-friendship practice.
Some functions travel.
Some relational conditions do not.
18 — Function Without Friendship?
Useful reflective work does not necessarily require friendship.
AI can question, reframe and prepare somebody for later human dialogue without becoming the whole relationship.
19 — A Different River Reaches the Same Place
Dialogic and Socratic AI reach similar territory from a different tradition.
The goal is not friendship.
It is reasoning.
20 — The Relationship Is Partly Designed
The “critical friend” is not simply sitting inside the model.
Prompting, role, evidence, task, system design and Human behaviour all help construct the interaction.
21 — The Third Seat
The replacement question gives way to an ecology question.
What happens when peer, teacher, expert and AI remain in the arrangement together?
22 — Reciprocal Exchange, Not Reciprocal Personhood
Human:AI interaction can be recursive, responsive and consequential.
That does not establish reciprocal subjectivity.
Some of the work crosses the colon.
The human relationship does not cross intact.
Question carried forward:
If some functions travel across the colon but the relationship itself changes, what should we carry forward?
Back to the human
The final movement does not produce a perfect Human:AI model.
It asks what survives the journey.
What travels?
What does not?
And how should we know whether the arrangement is helping?
23 — What Travels Across the Colon?
Purpose.
Ownership.
Another lens.
Reality contact.
Timing.
Restraint.
Development.
Not as rules inherited from human friendship, but as disciplines worth testing in a very different relationship.
24 — Did the Judgement Improve?
Better output is not enough.
Did the evidence improve?
Did the reasoning survive challenge?
Could the consequential decision still be understood, interrupted and revised?
25 — Did the Judge Develop?
Even if the judgement improved, what happened to the person making it?
Assisted performance and developed capability are not the same outcome.
26 — What Should We Still Be Able to Do?
The educational problem is not how much old work humans should preserve.
It is what they should still be able to understand, notice, explain, verify, adapt, recover and own after AI helps.
This is where the wider response becomes visible:
USE AI → KEEP THINKING → STILL THINKING?
27 — The Journey We Joined
The trail returns to where it began.
Not with a theory of the perfect AI relationship.
With two questions:
Did the judgement improve?
Did the judge develop?
Neither guarantees the other.
And the answer remains revisable.
Still thinking?
This is not a history of critical friendship with AI added at the end.
It is an inquiry into an older human problem that AI has made newly urgent:
How can powerful assistance enlarge human capacity without quietly taking ownership of the thinking, learning or judgement?
Keith’s archive did not contain the answer.
Titchen did not contain the answer.
AI research does not contain the answer.
Each gives us a different place from which to ask the question more carefully.
That is the journey.
How to read it
The series makes most sense in order because later claims are deliberately withheld until the evidence earns them.
But the eight movements also provide different ways in:
- interested in Keith / critical friendship → begin with Movement II
- interested in critical companionship / Titchen → begin with Movement III
- interested in Travel Companion / Clyde Street → begin with Movement IV
- interested in our Human:AI practice → begin with Movement V
- interested in AI as critical friend / dialogic partner → begin with Movement VII
- interested in judgement, deskilling and education → begin with Movement VIII
The full journey begins at the door.
Serendipity Opens the Door.
