The Journey We Joined

What would make us revise even this?

The same things that have governed the journey from the beginning: evidence, reality, and what happens to the person.

To see why, return to the door.

The series began with an accident.

A folder opened.

Inside it was a trail of writing about critical friendship that Keith Lyons had collected years earlier.

The temptation was immediate.

The questions looked familiar.

Challenge.

Support.

Ownership.

Trust.

Independence.

Another way of seeing.

It would have been easy to turn resemblance into inheritance.

We did not.

We stayed in the folder.

That changed things.

Critical friendship turned out not to be one idea.

It contained disagreements about who the friend served, how authority entered the relationship, whether people actually knew how to challenge one another, and whether a useful professional exchange deserved to be called friendship at all.

Then one citation led out.

Gibbs and Angelides pointed to Angie Titchen.

Critical companionship.

Again, the temptation was obvious.

Keith would later call himself a Travel Companion.

But there was no evidence that Keith had followed that citation.

So we followed Titchen without him.

Her work gave companion more weight.

Mutuality.

Reciprocity.

Particularity.

Timing.

High challenge and high support.

A journey undertaken together.

And eventually, parting.

Only then did we return to Clyde Street.

Keith’s Travel Companion was not Titchen’s critical companionship.

But the resemblance became more interesting once lineage was removed from the story.

Keith had spent years changing position around coaches.

Leading sometimes.

Following sometimes.

Listening when challenge would have been wrong.

Remaining available when people chose not to lean.

Being corrected.

Learning too.

The relationship had become less about delivering expertise and more about staying close enough to another person’s journey to help without owning it.

Then the direction changed again.

Keith had not only accompanied others.

In our own history, learning sometimes travelled back towards him.

The relationship was not symmetrical.

But it was not one-way.

That mattered when we finally placed the tradition beside Human:AI.

Much of the shape was already there.

Before the old folder reappeared, our working practice had developed a human Chair.

Contribution separated from authority.

Challenge without domination.

Rollback.

Reality contact.

Counter-routes.

Method-as-object.

A concern for what the human continued to understand, notice, practise and own.

The resemblance was real.

The family tree was not.

And the differences mattered just as much.

AI could contribute on a scale no human critical friend could.

Retrieve across huge evidence fields.

Generate alternatives instantly.

Hold complex working states.

Challenge at any hour.

Draft, compare, code, test and revise.

But it did not arrive with a human life.

No independent experience.

No vulnerability of its own.

No moral concern because something had happened to it.

No reciprocal subjectivity merely because the exchange felt reciprocal.

The colon became useful precisely because the participants were unequal.

Human:AI

A relation.

Not an equivalence.

Then the literature itself moved into the AI era.

Researchers began asking whether AI could act as a critical friend.

Some functions transferred.

Questioning.

Feedback.

Alternative perspectives.

Structured reflection.

Socratic opposition.

AI could become useful before a human conversation, alongside one, or inside a wider ecology of peer, teacher and machine.

But the emerging evidence kept returning to the arrangement.

Prompting mattered.

Role design mattered.

The surrounding people mattered.

What the AI could see mattered.

What the human still did mattered.

The relationship did not live inside the model alone.

And that brought us back to the question that existed before this series started:

The work got better. What happened to the human?

We can now answer more carefully.

Sometimes the human got better too.

Sometimes AI released attention for more important work.

Sometimes it exposed an assumption, widened the evidence, sharpened the judgement or gave the person a route into something they could not previously do.

Sometimes the work improved while the human did less of the cognitive activity through which capability had once developed.

Sometimes nobody could tell from the final artefact which of those things had happened.

So better work is not enough evidence.

Neither is worse work.

Neither is AI use.

Neither is human presence.

The series has left us with two questions that need to stay separate:

Did the judgement improve?

and:

Did the judge develop?

A good Human:AI arrangement may improve one, the other, both or neither.

The important thing is not to infer one from the other.

That may be where the old critical-friend question reaches the AI era most cleanly.

The helper should not become the owner of the learning.

The help should not make its own continued presence the only route to competent action.

And the person should not remain in the loop merely as decoration.

Human judgement has to remain usable.

Purposeful.

Informed.

Challengeable.

Revisable.

Responsible.

Capable of learning.

That is not something AI can preserve for us automatically.

It is something the arrangement has to make possible.

Which is why the final return is not a model of the perfect Human:AI relationship.

There is no context-free version.

The right arrangement for an expert doing routine analysis may be wrong for a novice forming expertise.

The right arrangement for reversible drafting may be wrong for a consequential decision.

The right arrangement today may become wrong when the system changes.

So the journey does not end with:

Here is how to use AI.

It ends somewhere closer to the way it began.

With a question.

What is the work for?

What can AI genuinely add?

What does the human still need to understand, practise, judge and own?

What evidence tells us the arrangement is working?

What would make us change it?

And what happens to the person over time?

That is the journey we joined.

Keith’s archive did not give us the answer.

It gave us older versions of the problem.

Titchen showed another way the problem had been lived.

Keith’s later practice showed what accompaniment could look like without a neat definition.

Our own Human:AI work arrived at related disciplines through a route whose provenance was mixed: some practices were older, while some later language was already being shaped inside a Keith-informed project.

Later research then placed AI directly inside the question.

The lines meet.

They do not collapse.

Perhaps that is enough.

Serendipity opened the door.

Curiosity took us through it.

Evidence limited how far we could claim to have travelled.

Sensemaking brought the journey back to the work.

And now the question returns to where it belongs.

Not to the AI.

Not to the archive.

Not even to the framework.

To the next judgement.

Still thinking?