Students need our help. They are kids.
That statement may sound obvious, but it sometimes seems strangely absent from our AI conversations. We can tend to speak about students as though they are miniature adults making fully informed ‘grown up’ decisions about authorship, cognitive offloading, intellectual honesty, privacy, bias, data, assessment policy, and disciplinary thinking.
They are not.
Even if our K-12 students were operating as adults, let’s remember the reality of what’s also happening in the adult world. Many adults, too, are struggling to make wise decisions about AI use. Professionals in law, business, academia, media, and education are all trying to work out where assistance ends and substitution begins.
K-12 students will, obviously, need guidance even more than adults. But are they truly getting that guidance?
Too often, schools offer either a ban or a blank cheque. Neither is sufficient.
A ban may be appropriate in particular contexts. There are times when students need to work without AI: memory, retrieval, handwriting, foundational knowledge, first attempts, in-class writing, oral explanation, source analysis, and controlled demonstrations of learning.
But a blanket ban does not teach wise use.
A blank cheque is worse. It leaves students to work out the ethical and cognitive boundaries of AI by themselves.
We need something clearer.
A quick recap of the story so far…
In Part 1 of this series, I argued that our response to AI in assessment cannot simply be framed around cheating. Yes, assessment integrity matters. Yes, some students misuse AI. Yes, schools need clear boundaries.
But if we make the conversation only about detection, punishment, or compliance, I think we miss something deeper. I argued that AI misuse is not only an integrity problem. It is also a teaching and learning problem. It is a motivation problem. A confidence problem. A task-design problem. A metacognition problem. A culture problem.
And, in a very human sense, it is a problem of pressure.
Students are growing up in a world where the boundaries between help, collaboration, automation, and authorship are becoming harder to see. Many adults are struggling with these boundaries too. So, yes, students need boundaries. But they also need guidance. They need a language. They need routines. They need modelling.
They need adults who can help them understand when AI supports learning and when it quietly hollows it out.
In Part 2, I began unpacking the first of eight possible responses: teaching students the difference between productive struggle and wasted struggle.
In Part 3, I explored the importance of redesigning tasks for an AI age and the need to place value on ‘the climb’ by making the process visible and valued.
This post, Part 4, picks up on another major challenge:
We need to make expectations around AI use clear.
That sounds simple. It is not.
Bill Walsh’s “base camps” and the Standard of Performance
This is where I return again to Bill Walsh. In The Score Takes Care of Itself, Walsh does not simply tell people to “do better”. He does not merely demand success. He hard works to define the behaviours, actions, attitudes, expectations, and cultural norms that make success possible. That matters in our classrooms too.
There is a section in the book where Walsh writes about establishing a Standard of Performance. In the pages I have been revisiting recently, he talks about identifying the specific actions and attitudes relevant to performance and production.
Start with a comprehensive recognition of, reverence for, and identification of the specific actions and attitudes relevant to your team’s performance…
Be clarion clear in communicating your expectation of high effort and execution of (this) standard of performance.
Like water, many decent individuals will seek lower ground if left to their own inclinations. In most cases you are the one who inspires and demands they go upwards rather than settle for the comfort of doing what comes easily…
(Walsh, 2009, pages 28-29)
That section has stayed with me.
Specific actions. Specific attitudes.
Not vibes. Kids won’t figure out what we mean.
Not “use AI responsibly” written vaguely in a document that students never read and teachers interpret differently.
“Specific actions and attitudes”
We need to be able to identify, define, clearly (and compelling) explain what “responsible use” means.
Students need to know not only what they may or may not do with AI in a particular situation. They need to know why. And they need to own the why.
They need to understand the beliefs underneath the boundaries in the many different given contexts where AI might be used. Walsh writes about teaching your beliefs, values, and philosophy. That, too, matters for schools.
Realistically, AI expectations cannot be some definitive checklist of ‘dos and don’ts’.
As a general purpose technology, AI has an enormous range of possible use cases by a diverse group of people in a variety of contexts. The expectations teachers set must be “clarion clear” yet flexible enough to help us to navigate these places of complexity and contestability.
Therefore, our expectations must be grounded in an understanding of the nature of AI. They must be grounded in an understanding of what we are trying to achieve in our learning processes.
They must be grounded in values: ethical, pedagogical, and cultural.
Students need to understand that the point of boundaries is not institutional suspicion. The point is to support their growth as people. The point is their agency. The point is their voice. The point is their capacity to do hard things in a world that increasingly offers easy substitutes.
Walsh contends that in high performance cultures we need to establish a metaphorical ‘base camp near the summit of achievement’. Setting clear expectations and high expectations was, to Walsh, a foundation on which to build ongoing achievement.
[This foundation was a “benchmark” – a way of] enabling us to establish a near-permanent base camp near the summit, consistently close to the top…
[To stay at this base camp of expectations, Walsh “had to”] teach everyone what they needed to know to get to where I wanted us to go.
(Walsh, 2009, page 27).
That connects beautifully to Miley’s lyrics in The Climb. It ties in with an understanding of productive struggle. It links clearly with having clear and visible goals.
Students do not need us to pretend the mountain is easy. They need us to help them establish base camps.
Places of stability and clarity in a VUCA world of AI and learning. Places of shared expectation. Places where the next stage of the climb becomes possible.
Clear AI expectations build those base camps. They are not the summit. Not the whole journey. But a stable place to return to when things wobble.
Clear expectations are not the same as overly-simplistic rules
I am increasingly convinced that one of the weakest responses to AI in schools is the over-simplistic rule.
“AI is banned.”
“AI is always allowed.”
“Everyone must use pen and paper for this!”
“This is a green level AI assessment task.”
“This is a Level 3 AI activity.”
“AI is cheating.”
These statements are too blunt for the messy middle of classroom life. They may offer temporary clarity and comfort to adults, but they rarely help students make wise decisions in messiness of real learning situations. Teachers need to wade into that ‘messy middle’ – the volatile, uncertain, complex and ambiguous space – of working with AI alongside their students.
The students in our classrooms are at a diverse range of ‘positions’ in their learning. A student drafting a paragraph is not in the same position as a student trying to understand a difficult concept. A student encountering a source for the first time is not in the same position as a student seeking feedback on a second draft. A Year 8 student in the early stages of historical source analysis is not in the same position as a senior student pressure-testing a historical argument. A novice learner is not in the same position as a more proficient learner. A learner may be a novice in one aspect of their work and an expert in another aspect – at the same time!
We need to reconceptualise what it means to be teaching with AI as a relational presence in our classes – and then think about what our expectations actually are for our learners.
This is why my Bubble and Burner model has become so useful to my thinking. At its heart, the model asks teachers to think carefully about how, when, and why AI might be used by diverse individuals within the sequence of their process of learning.
The Bubble and Burner model enables teachers to reflect on their expectations around AI use by the diverse learners in their classrooms – who are all at different positions within the process of learning.
So many questions come to mind about our expectations around AI use in our classes.
When should AI be withheld? Who for? Why? For how long?
When should AI be bounded? In what way or ways? How is this communicated to students?
When should AI be introduced? For the class? For some of the class? For an individual?
When should the teacher step in and turn the heat down — or even activate the “emergency off”?
I’m not even going to pretend that I can give you an answer for every classroom. My answers lie in the need for teachers to deeply reflect on the realities of an AI-infused classroom – and to reconceptualise the teaching and learning experience within it.
The Bubble and Burner model positions the teacher as the regulator of the learning process, adjusting the “flame” of challenge and AI use so that students experience enough cognitive friction to learn, but not so much that they are overwhelmed. That is the work. And it’s hard. It’s context specific. It’s activity specific. It’s phase of learning specific. It’s learning needs specific.
It’s messy.
It is not a neat traffic-light poster.
It is professional judgement.
It requires thoughtful consideration of what expectations are required within each and every part of the learning process. It requires teachers’ careful consideration of how to make those expectations “clarion clear” within the context of their students’ learning journey.
From “Can I use AI?” to “What kind of help is appropriate here?”
Students often want a simple answer:
“Can I use AI for this?”
That is understandable.
But I think we need to teach a better question:
What kind of AI help is appropriate at this point in my learning?
That shift matters.
It moves the conversation away from permission and towards judgement.
It helps students see that AI use is not one thing.
There are different modes of AI use, and they have different levels of risk, value, and suitability. When setting clear expectations around AI use, perhaps it’s useful to conceptualise student learning with AI via a framework gives students some boundaries without pretending that indiviualised context specific AI use does not exist. What matter is how much of the productive struggle of learning has the student engaged in. Have they done ‘enough thinking first up’ before they are seeking to use AI?
For example:
- No AI might be best for novice learners: for memory, retrieval, foundational skills, handwriting, first attempts, and controlled evidence of learning.
- AI as explainer might be a good way to support the accessibility needs of students: to clarify a concept, define terms, or provide examples.
- AI as coach might be good as a student’s expertise develops through revision: to ask questions, offer hints, quiz, to guide understanding of criteria rubrics, or identify gaps.
- AI as critic might be good when a student is operating as expert editor: to challenge a draft, identify weaknesses, or suggest counterarguments.
- AI as collaborator or even co-producer might be helpful when a student has a high level of skill: to compare interpretations, refine structure, or generate alternatives.
What’s core, again, is teaching students to ask the following question of themselves and to truly own the honest answer:
What kind of AI help is appropriate at this point in my learning?
Teachers need to help students to discern how and when to use AI best to support their own learning. Teachers need to teach students how to make the process of learning with AI transparent. Teachers need to ensure their expectations demand that the student remains fully accountable for both the process and product of their learning.
Such an approach recognises the reality that students in the same class may be at different points in ‘the climb’ at the same time.
A student with strong established knowledge of an aspect of a topic may be ready to use AI as a critic. Another student in the same room may still need to focus on retrieval, vocabulary, and first-draft thinking before AI enters the process. A third might gain great value through using AI to access the task!
This is why teacher judgement remains indispensable.
Traffic-light systems and AI scales may help guide our thinking. But they cannot replace knowing our learners and how they learn.
Walsh’s Standard of Performance was precise. It was behavioural. It was taught. It was clarion clear. Ours should be too.
BUT ours will need to be flexible. Context specific, Build upon values. Built upon wisdom.
The classroom is NOT a football field.
Our standard of performance must help our students navigate the volatile, uncertain, complex and ambiguous space of learning with AI as a relational presence in the classroom.
Our standard of performance demands judgements – by teachers and students.
What kind of AI help is appropriate at this point in the learning process?
A quick dive into the bubbles: Novices and experts
In the Bubble and Burner model, I describe AI use through the metaphor of “bubbles”. Some bubbles are small and low. Some are large and high.
Small-low AI use cases might include asking AI to define a word, clarify instructions, suggest a study question, explain a concept in simpler language, or help a student identify what they do not yet understand. These uses can support access without stealing the learning. They can be examples of beneficial cognitive offloading.
Large-high use cases are different. They might include asking AI to critique an argument, simulate alternative interpretations, generate counterarguments, refine a complex structure, or act as a debate partner. These uses can be powerful. But they require more expertise.
A student who has not yet built enough knowledge may not be able to judge whether the AI’s response is shallow, misleading, overconfident, or just plain wrong. If a student chooses a large-high use case when they haven’t met the challenge of the climb, they risk the appearance of competence because they have skipped a requisite amount of productive struggle.
This is the novice-expert issue. The “small-low” and “large-high” problem
Experts can offload differently because they already have knowledge structures to protect them. Novices cannot safely outsource what they have not yet built.
This is why “Think First” is a core expectation of the base camp.
Before students move into more complex AI use, they need to have done enough human thinking to remain responsible for the journey.
Think First: The Monster Expectation of the Base Camp
In my post A Tale of Two Monsters: Thinking First in an AI Age, I described two “little monsters” I use as visual cues in my classroom: Think First / Train Your Eyes and Bee Discerning / ADAMANT RUP. They are playful, but their purpose is serious. They function as memory anchors for attention and judgement.
The first monster says:
Do not outsource your first encounter with evidence.
The second says:
Be discerning. Analyse carefully. Judge thoughtfully.
These are not decorative classroom mascots. They are part of a Standard of Performance. They help students know what kind of thinking is expected at different moments.
Before AI, students must look.
They must notice.
They must wonder.
They must infer.
They must struggle a little with the evidence.
Then, once they have done that thinking, AI may enter as support for their learning – not as substitute for productive struggle.
This is where the Little Monsters connect directly to the Bubble and Burner model. In that post, I described the key question for teachers as not simply whether students should use AI, but how we sequence and scaffold help so that students remain the primary meaning-makers.
That is the base camp.
Attention before assistance.
Discernment before dependence.
Human judgement before machine fluency.
Our standard of performance in an AI-infused classroom requires constant judgements. It must be robust enough to support students as they navigate the volatile, uncertain, complex and ambiguous space of learning with AI as a relational presence. Our standard of performance demands that we constant check in with the question:
What kind of AI help is appropriate at this point in the learning process?
Prompting routines: Expectations of Base Camp
While it might be tempting to buy into the language of ‘master prompts’ and ‘building prompt libraries’, these do little to teach the complexity and judgement that students (and adults) need to navigate their VUCA AI world. We need to demand more of them in the ways that they work with AI as a tool.
This is also where prompting routines matter.
I have developed various routines over time. The 2 Is and a Verb routine. The Plus 3 move. The CEC approach.
These routines are not magic spells. No acronym is.
But they can help build stable base camps for students who otherwise may not know how to begin.
In an AI-infused classroom, prompting is not merely a technical skill. It is a thinking skill. A poor prompt often reflects unclear thinking. A better prompt often requires the student to know:
- what they are trying to achieve
- what role they want AI to play
- what information AI needs
- what constraints matter
- what kind of response would be useful
- how the output will be judged
That is not just “using AI”. That is metacognition.
So, a routine like 2 Is and a Verb can help students slow down, to think first, and develop their prompting skills.
For example, before using AI, students might need to identify:
- Intent — What am I trying to learn or improve?
- Information — What context or material do I need to provide?
- Verb — What do I want AI to do?
Explain. Quiz. Challenge. Compare. Clarify. Critique. Summarise. Suggest.
Each verb changes the learning relationship.
“Write this for me” is very different from “ask me three questions that help me improve my argument”.
The Plus 3 move can then push students beyond the first easy answer. It encourages productive struggle. It encourages a cognitive friction while using AI. It requires students to move beyond the first response from an AI tool. The point is thinking.
The CEC approach helps in this stage. It sets a clear expectation that students Clarify the responses it gets from AI; that they demand Elaboration, further Explanations, or detailed Evidence; and that the Challenge the responses – that they Boss the Bot and Call Out The BS. (You may like to read about these in more detail on my blog OR in my peer-reviewed journal article HERE.)
Teaching beliefs, values, and philosophy
Walsh’s reminder about teaching beliefs, values, and philosophy is important. In schools, we often try to solve cultural problems with procedural documents. Procedures matter. Policies matter. But students need more than rules.
They need to know what we believe learning is. They need to know why thinking matters. They need to know why their voice matters.
They need to know why integrity is not merely about avoiding punishment. They need to now that achievement isn’t just about grades. They need to know that help is good, but help that replaces growth is not.
They need to know that AI can be a useful study buddy, critic, coach, or explainer, but it must not become the hidden author of their thinking. They need to know that it’s ok to struggle and to make mistakes. That a polished answer is not the same as a changed learner.
We should teach these ideas explicitly.
Not just once. Not just ‘done’ in a hot take on an assembly. Not just performatively buried in an assessment cover sheet or in a student diary – never to be read.
Taught repeatedly.
In the language of the classroom.
In the routines of lessons.
In the feedback we give.
In the questions we ask.
In the way we respond when students make mistakes.
This is culture work. This is slow teaching. This is how the Standard of Performance builds a base camp for learning.
A Base Camp Pledge? A possible classroom standard of performance
Many teachers and school leaders are searching for a set of rules around AI use. As this post will likely come out during the first week of semester 2 for many Australian teachers, perhaps this is an offering of a suitable ‘rule’.
I wonder whether a classroom AI Standard of Performance might sound something like this:
In this classroom, we think first.
We use Human Intelligence before Artificial Intelligence.
We make our thinking visible. We work transparently.
We use AI as support for thinking, not as substitute for it.
We ask for help when struggle becomes wasted, not simply because learning feels hard.
We remain responsible for our thoughts and words.
We show gratitude and acknowledge assistance freely and honestly.
We can explain what changed in our thinking.
We value ‘the climb’ – not just the ‘summit’.
That is not a policy. It is a statement of a culture. It is a base camp.
It gives students somewhere to stand when the journey is wobbly.
Clarion clear expectations are an act of care
There is a temptation to see clear AI expectations as restrictive. I think they are an act of care. Students deserve to know where they stand. They deserve to know what kind of help is appropriate. They deserve to know when AI use is wise, when it is risky, and when it undermines the learning they are meant to do.
They deserve teachers who do not simply say, “Don’t use it.” They also deserve teachers who do not simply say, “Go for it.” They deserve adults who will teach them the difference. That is demanding work. But good teaching has always been demanding.
Walsh understood that high standards are not the enemy of care. In fact, when they are communicated well, modelled consistently, and connected to growth, they are a form of care.
The same is true in classrooms.
We are not setting AI expectations because we distrust students.
We are setting them because we believe students are capable of becoming more thoughtful, more discerning, more independent, and more fully human learners.
Clear expectations matter.
They help establish base camps from which to achieve new heights.
They give students safe, predictable, and structured places to return to in a volatile, uncertain, complex and ambiguous world of schools when the pressures rises, when the assessment feels too hard or too much, when the shortcut is tempting, when the AI offers convenient relief.
A good base camp makes the climb possible.
In an AI-disrupted classroom, clear expectations are not merely administrative. They are pedagogical. They are ethical. They are relational.
They tell students:
You are allowed to seek help.
You are expected to think.
You are responsible for your judgement.
You do not have to climb alone.
But the climb still belongs to you.
And perhaps that is the work before us.
Not simply to guard the scoreboard.
Not simply to police the summit.
But to build classroom cultures where students know the standard of performance well enough to keep climbing with honesty, courage, discernment, and agency.
Bonus blog post: ‘Clarion Clear’ Expectations and ‘Base Camps’: Some practical classroom moves for teaching with AI
Disrupted History 
You must be logged in to post a comment.