How do we know whether a critical friendship actually did the work we are claiming for it?
By 2024, Elizabeth Petroelje Stolle and Charlotte Frambaugh-Kritzer had been living with versions of that question for more than a decade.
They had moved from trying to define critical friendship, to accepting its variability, to building a continuum that made different configurations easier to describe.
Then they turned the question back on their own research.
Their study Critical Friendship Quality and Trustworthiness: Our New Understandings examined eight of their earlier self-studies across ten years.
The object was no longer simply what kind of critical friendship did we have?
It was closer to:
What makes us entitled to claim that this critical friendship contributed quality or trustworthiness to the research?
That is a much harder question.
Critical friends have often appeared in research methods sections as reassuring figures.
They challenged assumptions.
Improved trustworthiness.
Provided another perspective.
The verbs sound useful.
But unless the study shows what happened, they can become methodological decorations.
A reader is asked to believe that critique occurred because the relationship was named.
Stolle and Frambaugh-Kritzer's later work pushes in the opposite direction.
They use a Critical Friend Quality Criteria Guide alongside the Definition Continuum and a catalogue of trustworthiness indicators. Their argument is that more detailed, shared language can make the critical friend's contribution clearer and more credible to the reader.
In other words, the relationship itself becomes something that has to leave evidence.
Not evidence that two people liked working together.
Evidence of critical work.
Was an alternative perspective actually introduced?
Did an assumption move?
Was an interpretation reframed?
Did the inquiry become more trustworthy in a way that can be explained rather than asserted?
The journey from 2018 to this point is worth noticing.
At first, the problem was that the term critical friend was being used too loosely.
Then the continuum made its variability visible.
Now the question has shifted again:
visible variation is still not enough.
The relationship has to be accountable for what it contributes.
That does not mean every critical friendship needs a bureaucratic scoring system.
There is a danger here too.
The more seriously a field takes a complex relationship, the more tempting it becomes to turn that complexity into criteria, forms and procedural compliance.
A box can be ticked for 'alternative perspective' without anybody's judgement being disturbed.
A protocol can document critique without making the critique useful.
We have seen that problem before.
A role can be agreed without being enacted.
A relationship can be called critical while remaining comfortable.
A method can become more visible without becoming better.
The later literature does not remove that tension.
It makes the tension inspectable.
The 2025 book Critical Friendship as a Self-Study Research Tool: A Comprehensive Resource Exploring the Complexities is evidence of how far the question had travelled.
Its chapters move through definition, meaningful connections, selection, invitation, working in relationship and assessing impact.
That structure tells its own story.
The field had not reached the point where critical friend could simply be dropped into a methods section and allowed to explain itself.
Twenty years of use had produced enough variation and ambiguity to justify an entire book about how to enact and account for the relationship.
This gives us a useful checkpoint before the next movement.
The post-2016 literature has made at least three things harder to ignore.
First, there is no single critical-friend configuration.
Second, the title tells us very little without enactment.
Third, the claimed value of the relationship should be inspectable rather than ceremonial.
Those lessons matter because the next part of this journey introduces a new candidate for the role.
Artificial intelligence.
And AI makes it extraordinarily easy to satisfy the surface features.
It can respond instantly.
Ask questions.
Offer another perspective.
Critique a draft.
Suggest alternatives.
Reframe an argument.
Sound encouraging while doing it.
If we had arrived here directly from Costa and Kallick's 1993 description, that might have looked like enough.
The intervening literature has made the bar much higher.
Before calling AI a critical friend, we now know to ask:
What is the configuration?
What is the purpose?
What kind of challenge is occurring?
What is missing from the relationship?
What evidence shows that the interaction changed anything?
And what does the word friend smuggle in that the system cannot actually provide?
Only now does the next experiment become worth following.
What happens when the critical friend becomes AI?
References
Petroelje Stolle, E. and Frambaugh-Kritzer, C. (2025; published online 2024) 'Critical Friendship Quality and Trustworthiness: Our New Understandings', Studying Teacher Education, 21(1), pp. 22–38. doi:10.1080/17425964.2024.2443572. Publisher article.
Stolle, E.P. and Frambaugh-Kritzer, C. (2025) Critical Friendship as a Self-Study Research Tool: A Comprehensive Resource Exploring the Complexities. Cham: Springer. doi:10.1007/978-3-031-91022-7. Publisher record.