Good teaching in the age of AI
When answers become abundant, getting the answer is no longer reliable evidence that learning has occurred.
That was the question beneath a recent workshop I facilitated with tutors at Y Education in Christchurch.
The session was not a tour of new AI tools. It was not about writing better prompts, and it was not primarily concerned with detecting whether learners had used ChatGPT.
Instead, we began with a more fundamental educational question:
What thinking did the learner still need to do?
That question shifts the conversation away from treating AI as a separate technical problem and returns us to the deeper work of teaching: designing learning through which people develop judgement, capability, confidence and independence.
One of the clearest messages from the room was that the relationship between education and AI remains unresolved—and may not need to be resolved before good teaching can continue.
We do not need a single position on AI
Not every kaiako wants to use AI.
Not every ākonga wants to use it either.
Some people are enthusiastic. Others are interested but cautious. Some are largely neutral. Others would rather the technology did not exist at all.
Those positions are already present across our classrooms, organisations and communities.
And that is okay.
A useful learning framework cannot depend on everyone becoming enthusiastic about a particular technology. It needs to work whether AI is actively used, quietly present in the background or deliberately removed from an activity.
Strengthening learning works with AI in the room.
It also works when AI is taken out of the picture.
That is because this is not fundamentally an AI framework.
It is a framework for good teaching.
AI simply makes the underlying educational questions more urgent.
Answers have become abundant
For most of human history, finding a good answer required effort.
You might need to ask a teacher, find a book, consult someone with experience, compare several sources or attempt the problem yourself.
Today, almost every learner carries access to an intelligent assistant. They can generate explanations, summaries, examples, images, code, feedback and completed written responses within seconds.
The cost of obtaining an answer has collapsed.
This does not mean knowledge no longer matters. Nor does it mean educators are becoming less important.
It means the opposite.
When answers are easy to obtain, the ability to evaluate, question, interpret and apply them becomes more valuable.
The important capabilities increasingly lie in what learners do around an answer:
- What did they notice?
- What did they question?
- What evidence did they examine?
- What assumptions did they test?
- What decisions did they make?
- What did they reject?
- What changed their mind?
- Could they explain their reasoning?
- Could they apply the learning somewhere new?
These are not new educational concerns. Good educators have always cared about them.
Generative AI simply makes them harder to ignore.
A better question
Much of the current discussion begins with:
Has the learner used AI?
It is an understandable question, but it is incomplete.
A learner could complete an entire task without using AI and learn very little.
Another learner could use AI as one source among several, critically evaluate its output, revise their assumptions and deepen their understanding.
Knowing whether AI was present does not, by itself, tell us whether meaningful learning occurred.
A more useful question is:
What thinking did the learner still need to do?
This is a question educators can design for.
Where does the learner have to make a decision?
Where do they encounter uncertainty?
Where must they explain themselves?
Where do they test whether something is true?
Where do they transfer what they have learned into action?
Once we begin asking those questions, AI becomes part of the learning environment rather than the centre of the conversation.
Five conditions for visible thinking
The workshop introduced a framework built around five connected conditions.
Attempt
Learning begins with a visible starting point.
An attempt does not need to be correct. It reveals prior knowledge, uncertainty, misconceptions and emerging understanding.
Before substantial assistance is introduced, what can we already see of the learner’s thinking?
Question
Strong learners do not only seek answers. They learn to ask questions that deepen understanding.
What are they really trying to understand?
What assumption are they making?
What question have they not yet asked?
When answers become abundant, the quality of the question matters even more.
Check
Information is not the same as knowledge, and confidence is not evidence.
Learners need opportunities to compare, test, verify and revise.
How do they know?
What evidence supports the answer?
What could be missing?
What would they do if the information proved unreliable?
Explain judgement
This is the intellectual centre of the framework.
The most valuable evidence of learning is often not the final answer, but the learner’s ability to explain how they arrived there.
A single question can reveal a great deal:
Walk me through your thinking.
What alternatives did they consider?
Why did they choose one approach over another?
What evidence influenced the decision?
What did they revise?
When judgement is explained, invisible thinking becomes visible.
Apply
Learning becomes capability when it can be used.
Can the learner adapt what they know to a new problem, workplace, context or challenge?
Can they do more than repeat an example?
Can they act with increasing independence?
Application is not simply the end of the process. It creates the experience from which the next attempt begins.
Together, these conditions form a cycle rather than a checklist. They provide educators with shared language for recognising and designing learning in which thinking remains visible.
Much of this is already happening
Another important observation emerged from the tutors themselves.
Making thinking visible does not necessarily require a new layer of activities, documentation or workload.
Kaiako already have many of these conversations.
They ask learners to explain what they are doing. They watch learners attempt practical tasks. They question a decision. They ask what went wrong. They invite learners to compare options, demonstrate a process, correct an error or try again.
These moments already reveal thinking.
The difficulty is that they may not always be recognised for what they are.
They may sit outside the formal assessment process. They may be treated as informal teaching interactions rather than useful evidence of reasoning, judgement and developing capability.
Some participants wondered whether this partly reflects the way organisations currently interpret NZQA requirements, particularly where assessment becomes narrowly associated with summative tasks and finished products.
That question requires careful examination. It should not be reduced to a claim about what NZQA does or does not require.
But it raises an important possibility:
The problem may not be that tutors need to do more.
The problem may be that we do not always recognise what existing practice already makes visible.
That does not mean every useful conversation must become an assessment event.
Nor should every interaction be documented, captured and bureaucratised.
The opportunity is to become more deliberate about the evidence already present in good teaching—and to decide when that evidence should inform professional or assessment judgement.
The vocational advantage
Trades and vocational education may hold a particular advantage here.
Vocational capability is often concrete, contextual and connected to real-world consequences.
Something was broken and it is still broken.
Or the learner fixed it.
A structure holds—or it does not.
A client receives a usable outcome—or they do not.
The learner produces a balanced espresso—or something fairly grim emerges from the machine.
The tools, materials, equipment and environment often provide immediate feedback.
That does not mean the final result alone proves understanding.
A learner might reproduce a successful process without being able to explain why it worked. They may complete one familiar task but struggle when the conditions change.
But when practical performance is combined with explanation, checking, adaptation and repeated application, vocational learning offers powerful evidence of capability.
The work itself talks back.
It provides consequences, resistance and feedback that are difficult to simulate through written answers alone.
This gives vocational educators a strong foundation for visible thinking:
- learners can attempt;
- the environment provides feedback;
- decisions can be discussed;
- errors can be diagnosed;
- judgement can be explained;
- learning can be applied again under changed conditions.
The more authentic the task, the harder it becomes for a plausible answer to substitute for actual capability.
From professional judgement to learning design
Frameworks become useful when they help us see something we could not see before.
During the workshop, participants worked with a series of short learning scenarios. Each scenario stopped before the outcome.
The purpose was not to decide whether the learner had behaved correctly or whether AI use was acceptable.
The central question remained:
What thinking remains visible?
Could participants see an attempt?
Was the learner asking meaningful questions?
Were ideas being checked?
Did the learner retain ownership of important decisions?
Could they explain their judgement?
Was there authentic application?
The discussions were particularly valuable where people disagreed.
Professional judgement is not strengthened by pretending that every situation has one obvious interpretation. It becomes stronger when educators make their reasoning visible, identify the evidence they are relying on and consider how another practitioner might reasonably interpret the same situation.
Participants were not only discussing visible thinking.
They were practising it.
The next step was to move from observing learning to designing it.
Using a learning-design canvas, participants selected one activity from their own practice and considered how it might be strengthened.
The challenge was deliberately contained:
Choose one activity.
Do not redesign an entire course.
They considered where learners would attempt, question, check, explain their judgement and apply what they had learned.
They also considered where AI might strengthen thinking without replacing it, and what evidence would demonstrate that learner thinking remained visible.
The Canvas was not intended as a compliance form.
It was a shared thinking surface.
The most useful question during the final discussion was not:
What did you write?
It was:
What changed?
Different groups were working with different activities, but many of their design moves were similar.
Learners would make an attempt before receiving substantial assistance.
Reasoning would be made explicit.
Information would need to be compared or verified.
Learners would explain why they made particular choices.
Application would move closer to an authentic context.
AI could remain present, but responsibility for judgement would stay with the learner.
What the workshop confirmed
The session reinforced several things for me.
First, educators do not necessarily need another list of AI tools.
They need useful ways of thinking about learning when AI is already part of the environment.
Second, there does not need to be complete agreement about AI before good educational work can proceed.
Kaiako and ākonga can hold different positions while still asking the same fundamental questions about learning, judgement and capability.
Third, much of the practice we need may already exist.
The task is not necessarily to add more activity, but to see existing conversations, demonstrations and professional judgements more clearly.
Fourth, vocational learning offers powerful models because capability is enacted in real settings and subjected to real feedback.
Finally, the most valuable resources do not remove uncertainty.
They help educators reason more clearly within it.
A continuing question
I left the workshop more convinced that dividing learning neatly into “AI” and “non-AI” activity will not take us very far.
Not every learner needs to use AI.
Not every educator needs to embrace it.
But all learners need opportunities to develop the thinking through which capability grows.
Are they attempting?
Are they questioning?
Are they checking?
Are they explaining their judgement?
Are they applying what they have learned?
When thinking is visible, educators can support it, challenge it, assess it and strengthen it.
When it is not visible, the quality of the final answer may tell us very little.
The framework, scenarios and learning-design tools will continue to evolve through further workshops, research and use in practice.
But the question underneath them will remain:
What thinking remains visible?
Because when we can see thinking clearly, we can design learning deliberately.