In June 2026, the Vancouver School Board rolled out Microsoft Copilot AI chatbot accounts for students aged 13 and up. Within days, a parent advocacy group called PACES Vancouver formed to push back. Within weeks, a national survey of Canadian teachers had quantified what educators were observing at the front lines.

Both responses – adoption and resistance – make sense. And the conversation they’ve generated reveals something more useful than either side alone: a clear articulation of what we actually want schools to develop in children, and a useful test for evaluating whether any school is building toward it.

What’s actually happening in BC schools right now

The Vancouver School Board’s Copilot rollout is the most visible local example of a province-wide shift. School districts across BC have been developing AI policies, approving AI platforms for student and teacher use, and navigating significant variation in how prepared they are to do so.

The BC Ministry of Education has published guidance documents on AI in K-12 schools, outlining considerations for responsible integration. But guidance documents and actual policy and training are different things. A Fraser Institute survey of 756 Canadian teachers in grades 6 through 12, conducted by Leger and published in June 2026, found that 64.7% of teachers had not been provided training or tools to identify when students are using AI to complete their work. Only 34.8% said their school has a policy regarding staff AI use, and only 42.3% said their school has a policy regarding student AI use.

The gap between the technology arriving in classrooms and the infrastructure to use it thoughtfully is substantial. AI tools are in BC schools now. Coherent, consistent frameworks for how students and teachers engage with them are, by most teachers’ accounts, still developing.

The parent advocacy group that formed in response to the VSB Copilot rollout, PACES Vancouver, has been clear about their concern: they’re not primarily worried about privacy or screen time. They’re worried about what happens to children’s ability to think independently when increasingly capable tools are available to do the thinking for them.

That worry has research behind it.

Why the parent concerns are legitimate, not just technophobia

The reaction to AI in classrooms has sometimes been characterized as technophobia – parents who are uncomfortable with change, who don’t understand the tools, who want their children’s education to look like their own. Some of that is probably true for some parents. But the specific concern about cognitive effects is grounded in a growing body of research, not just discomfort with the unfamiliar.

The mechanism researchers focus on is cognitive offloading: the transfer of mental effort to external tools. Cognitive offloading isn’t inherently problematic. Using a calculator for arithmetic frees mental resources for higher-order mathematical reasoning. Writing things down externalizes memory so working memory can be used for thinking. These are beneficial offloads because the underlying capacity remains exercised and develops independently.

The concern with AI tools is different in character. When students use AI to generate ideas they would otherwise generate themselves, to draft arguments they would otherwise construct, to find answers to questions they would otherwise have to reason through, they aren’t just offloading a routine task. They’re bypassing the effortful cognitive work that, in children and adolescents, is how thinking capacity develops.

A 2025 study by researcher Michael Gerlich, published in the journal Societies, examined AI tool use and cognitive skills across 666 participants and found a significant negative correlation between AI tool use and critical thinking ability, mediated by cognitive offloading. The study found this effect strongest in younger users. A 2025 MIT Media Lab study using electroencephalography found that participants who wrote essays using AI tools showed weaker neural connectivity during the task than those who used search engines or wrote unaided, and demonstrated lower sense of ownership over their work, with impaired recall of their own writing even afterward.

The researchers coined the term “cognitive debt” for this reduced neural engagement – the idea that using AI to bypass effortful thinking creates a deficit that compounds over time.

This research is not yet settled. Other studies find that AI can enhance thinking when used as a tool for elaboration rather than replacement – when it extends what students are doing rather than substituting for it. The distinction matters enormously, and it hinges on how AI is used rather than simply whether it is present.

But the Fraser Institute’s 2026 survey of Canadian teachers found that by their accounts, students are predominantly using AI to have their work done for them rather than to support their own thinking. Of students perceived to use AI, teachers estimated that 55.9% use it to do their work for them rather than to support it. 83.6% of surveyed teachers reported a decline in students’ critical thinking skills compared to five years ago. 72.9% reported that current students’ in-class writing and analytical skills are worse. These are Canadian teachers, describing Canadian students, in 2026. The policy options piece quoting the Canadian Teachers’ Federation survey of more than 6,200 educators found similar patterns – respondents warning that student use of AI is creating increasingly shallow learners who regurgitate information instead of critically thinking.

What the research says about learning and productive difficulty

The cognitive offloading concern connects to something researchers have understood about learning long before AI existed: difficulty is not the enemy of learning. In many cases, it is the mechanism of it.

Cognitive scientists describe this as desirable difficulty – the principle that learning tasks involving more cognitive effort, more struggle, more active processing, produce stronger retention and more transferable understanding than tasks made easy through scaffolding or shortcuts. Students who have to work to retrieve information retain it better than students given it directly. Students who have to construct arguments rather than have them generated retain the thinking process, not just the output.

This is why a student who uses AI to write an essay has not learned to write. They have produced writing. The same distinction applies to mathematical problem-solving, historical analysis, scientific reasoning, language learning, and virtually every domain where educators are worried about AI. The activity that produces learning is the effortful engagement, not the output that engagement produces. Remove the effort and you remove the learning, even as you improve the output.

A 2025 randomized study published in Proceedings of the National Academy of Sciences examined high school students using GPT-4-based math tutors. Students using the AI showed improved practice scores but performed worse on subsequent tests taken without the tool. The learning, to the extent it occurred, didn’t transfer. The performance did not reflect actual capability.

The question for families evaluating schools is not whether a school uses AI. It’s whether the school’s approach to learning – with or without AI – centers the effortful cognitive engagement that actually produces thinking capacity, or whether it optimizes for output at the expense of the process that develops real skill.

The skill that matters most when AI can do everything else

There’s a version of the AI-in-education debate that treats it as a question about cheating: will students use AI to bypass work they’re supposed to do themselves? This frames the issue as a compliance problem. Schools need better detection tools, stronger honor codes, more in-class assessment.

That framing misses the more fundamental question, which is not about cheating but about what schools are trying to develop.

If AI can write, calculate, analyze, code, and reason at sophisticated levels, the question is what human capacity remains distinctively valuable. The honest answer researchers are converging on is something like: the capacity to think originally, to evaluate critically, to know what question to ask, to recognize when output is wrong or limited, to synthesize across sources in ways that reflect genuine understanding, and to create in ways that reflect actual perspective rather than statistical aggregation of existing perspectives.

None of these capacities develop through passive consumption of competent AI output. They develop through the same thing they’ve always developed through: effortful, repeated practice of the underlying cognitive skills – reasoning, questioning, evaluating, constructing, revising.

The 2026 Gen(Z)AI Youth Assembly in Toronto, which brought together 100 participants aged 17 to 23, identified cognitive offloading and the consequences for their ability to learn, work, and engage in civic discourse as one of their three biggest worries about AI. This is not adults worried on behalf of young people. It is young people articulating their own concern about what AI dependence is doing to their thinking.

Why productive difficulty is the point, not the problem

The phrase “that’s hard” has been reframed in educational culture as a problem requiring a solution. When children struggle, the impulse is to reduce the struggle. When tasks take a long time, the impulse is to find ways to complete them faster. When answers are uncertain, the impulse is to provide certainty.

But productive difficulty – struggle within a student’s developing capacity, with appropriate support available when needed – is not a bug in educational design. It is the core mechanism. Children who are never allowed to struggle productively don’t develop the frustration tolerance, the persistence, the problem-solving strategies, or the genuine satisfaction of having figured something out that characterizes capable, confident learners.

Montessori education has been explicit about this for over a century. Maria Montessori observed that children who are given space to wrestle with challenges, who are not immediately rescued from difficulty, who are trusted to work through obstacles at their own pace, develop internal resources that children managed to easy success do not. The prepared environment isn’t designed to make everything easy. It’s designed to make everything appropriately challenging – accessible enough for genuine engagement, difficult enough for real development.

The 2023 Campbell Collaboration systematic review of Montessori research, which examined 32 rigorous studies, found consistent positive effects across all nine outcome measures including executive function, creativity, and academic achievement. The effect sizes were medium to large. Researchers identified the individualized, self-paced, hands-on engagement with materials – the productive difficulty baked into Montessori’s structure – as a key factor in those outcomes.

This approach didn’t emerge as a response to AI. But it turns out to be precisely what the AI moment calls for: an educational structure built around the development of genuine cognitive capacity through effortful engagement, rather than around the production of outputs that may or may not reflect real understanding.

What Montessori education has always known about genuine thinking

The Montessori approach to learning is in some ways an extended response to the question AI has now made urgent: how do you develop genuine thinking rather than the performance of it?

The answer embedded in Montessori’s design is that you develop it by structuring learning so that children have to do the actual cognitive work – choosing what to engage with, making sense of materials themselves, discovering patterns through direct experience, constructing understanding rather than receiving it. The teacher’s role is to observe, to introduce, to support when support is genuinely needed. Not to provide what children could discover themselves.

This is not inefficiency. It is the opposite of inefficiency: it is the most direct route to genuine competence, which requires the learner’s own cognitive engagement to develop. AI tools that substitute for that engagement are anti-educational in the deepest sense, not because they violate rules but because they bypass the very process that produces learning.

Our school’s approach to technology follows from these principles rather than from a specific policy about AI. We use technology where it serves learning – where it genuinely extends what students can discover, communicate, or create in ways that require and develop their own thinking. We don’t use technology in ways that substitute for the effortful engagement that develops cognitive capacity.

This is not a position against AI. AI will be part of our students’ professional and civic lives, and developing the ability to use it thoughtfully, critically, and selectively is genuinely important. But that capacity develops in students who have robust thinking skills of their own – who can recognize when AI output is wrong, limited, or misleading, who can evaluate rather than simply consume, who know what questions to ask and can tell the difference between a good answer and a plausible-sounding one.

Students who develop those capacities through years of effortful, self-directed, hands-on engagement with real materials and genuine challenges are better positioned to use AI as a tool than students who have outsourced cognitive work to AI throughout their education. The foundation matters. And the time to build it is before AI becomes the path of least resistance, not after.

What to look for in a school’s approach to technology and learning

For families evaluating schools with the AI question in mind, the relevant inquiry isn’t whether a school has an AI policy – most are still developing them. It’s whether the school’s fundamental approach to learning centers the effortful cognitive engagement that builds genuine thinking capacity.

Ask what a typical learning session looks like for a student in the program you’re considering. How much of what students do involves passive reception versus active construction? How much involves effortful problem-solving versus following provided procedures? How does the school respond when students struggle – is difficulty treated as something to eliminate or something to support students through?

Ask what the school is assessing. If assessment primarily measures outputs – essays, tests, answers – without evaluating the thinking process that produced them, the school is vulnerable to the AI displacement problem regardless of its official policy. If assessment involves direct observation of learning in progress, conversation between teachers and students about how they’re thinking, portfolio documentation of genuine development over time, the thinking process is being valued alongside the output.

Ask what the school’s learning environment does structurally. In a Montessori environment, the design of the prepared environment, the expectation of self-directed work, the multi-age structure, and the absence of competitive grading all direct students toward genuine engagement with learning rather than performance optimization. These structural features are harder to shortcut than any specific policy.

The AI conversation is ultimately a conversation about what learning is for. If it’s for producing outputs that demonstrate knowledge at moments of assessment, AI is a very effective shortcut. If it’s for developing genuine capacity to think, reason, question, create, and evaluate, then the shortcut misses the point entirely.

The concern BC parents have raised about AI in classrooms is legitimate and grounded in real research on what happens to thinking capacity when effortful cognitive work gets bypassed. The Fraser Institute’s June 2026 survey of 756 Canadian teachers and the Canadian Teachers’ Federation’s survey of more than 6,200 educators both document what those teachers are observing: students increasingly producing AI-generated work they didn’t engage with themselves, with measurable declines in critical thinking and analytical skills.

This isn’t a reason to panic or to dismiss AI as inherently harmful to education. It is a reason to ask pointed questions about what any school – independent or public, Montessori or conventional – is building in children. Whether effortful cognitive engagement is centered or bypassed. Whether difficulty is treated as something to support children through or something to eliminate. Whether what gets assessed is genuine thinking or polished output.

Curious what a learning environment built around genuine thinking looks like in practice? Book a campus tour at Westmont – we’d welcome the chance to show you.

Research Citations:

https://www.fraserinstitute.org/studies/survey-canadian-grades-6-12-teachers-ai-use

https://policyoptions.irpp.org/2026/08/ai-literacy-school-safeguards/