Is AI Helping Students Think, or Thinking for Them? When Technology Supports Learning and When It Replaces It
Caveat - During our high school presentations, we regularly ask students what they think about artificial intelligence and, perhaps more importantly, how they are actually using it in their education. Anecdotally, their answers are anything but uniform, some students openly tell us they use AI primarily as a shortcut, whether to complete an assignment, generate an answer, write something they do not want to write themselves, or simply get through schoolwork with as little effort as possible. For these students, AI can become less of a learning tool and more of a way to bypass some of the cognitive work the assignment was intended to develop.
Other students describe a very different relationship with AI, they are using it as a co-learning tool alongside the instruction they receive from their teachers. They may ask AI to explain a difficult concept in another way, provide an example that makes more sense to them, challenge an argument they are developing, quiz them before a test, provide feedback on something they have already written, or help them explore a topic beyond what could reasonably be covered during classroom time. For these students, AI is not necessarily replacing the learning process, it can become another tool they use to support and extend it.
There is also a third smaller group of students that is sometimes overlooked in this conversation, we have spoken with youth and teens who want little or nothing to do with generative AI. Some are uncomfortable with the technology, others have concerns about its accuracy, ethical implications, or environmental impact, while some simply prefer to learn, think, and complete their work without relying on AI.
What we find particularly interesting is that students from all three groups may be sitting side by side in the same classroom, working through the same curriculum, completing the same assignments, writing the same tests, and having their learning evaluated through many of the same traditional assessment methods. However, the way each student arrives at the finished product, the amount of cognitive work they perform themselves, and the role AI plays in that process may be significantly different. This raises an increasingly important educational question, “If the tools students use to learn are changing, how should we thoughtfully integrate them into teaching and learning, and do our traditional methods of assessing what students know and can do need to evolve alongside them?” This question sits at the heart of this article and the non-partisan educational discussions we believe need to take place with parents, caregivers, and educators collectively.
For the past two years, given our access to schools across Canada, we have witnessed how artificial intelligence is already changing how students research, write, study, brainstorm, solve problems, practise skills, receive feedback, and complete assignments. Whether parents, caregivers, or educators welcome that change, remain cautious about it, or would prefer generative AI tools to stay outside the classroom, there is a practical reality that schools increasingly have to confront, AI is already in the classroom, and its presence is only going to become more difficult to separate from the broader digital environments students use every day.
Even where particular AI tools are blocked on school networks or school issued devices, many students can still access AI through personal phones, home computers, browsers, search engines, productivity software, social media platforms, and an expanding number of apps where generative AI is simply built into the experience like Snapchat’s “My AI”. Increasingly, a student may not even think of themselves as deliberately “using AI” because the technology may be embedded inside a search engine, writing program, smartphone, educational platform, or other digital tool they already use.
This is one reason why we believe the educational conversation needs to move beyond the binary question of, “Do we allow students to use AI?”, and toward the much more useful question, “What should a student be able to understand, explain, evaluate, and do before, during, and after using AI?”
The emerging research is increasingly pointing toward an important distinction. The educational question is not simply whether students use artificial intelligence, but what role the AI is playing within the learning process. AI can sometimes act as a scaffold that helps a student understand difficult material, practise a skill, receive feedback, overcome a barrier, or explore an idea from another perspective. However, In other circumstances, the same technology can perform the very cognitive work the student was supposed to practice. Sometimes AI can be a scaffold, sometimes it can be a shortcut, sometimes it can function as a tutor, and sometimes it can become a substitute for thinking. Learning to recognize the difference may become one of the most important educational skills we can help students develop, as AI becomes increasingly integrated into their onlife world.
Producing an Answer Is Not the Same as Learning
One of the challenges with generative AI is that it can produce something that looks remarkably similar to learning. A student can ask an AI chatbot to explain a historical event, solve a mathematical problem, summarize a novel, generate an essay, create a thesis statement, write computer code, or answer a science question. Within seconds, the student may receive a polished, confident, and well structured response that looks every bit as sophisticated as something produced after considerable research and thought. However, the existence of a good looking answer does not necessarily tell us what happened inside the student’s mind.
Could they understand the material?
Could they explain it without the AI?
Could they identify an error in the AI’s response?
Could they defend the conclusion?
Could they explain why one source is stronger than another?
Could they apply the concept to a new problem?
Could they recognize when the AI was confidently wrong?
These are very different measures of learning than simply asking whether a student successfully produced an assignment. This distinction is becoming increasingly important as researchers examine how students actually use generative AI. The OECD’s PISA 2025 results provide an interesting example. Across OECD countries, approximately one in five students reported using AI tools almost every day. However, the relationship between AI use and academic performance was not straightforward. Students who reported using chatbots for particular schoolwork tasks often performed worse than students who did not. Yet students who frequently used AI specifically for learning and who also reported being taught how to evaluate AI generated information performed slightly better in science than students who rarely or never used AI. The OECD also cautioned that socioeconomic differences complicate these comparisons, with more advantaged students more likely to report receiving opportunities to learn how to evaluate AI generated information (1).
We think PISA tells us something important. The educational conversation should probably not be reduced to “AI use” versus “AI non-use.” Two students can both say that they used artificial intelligence to complete an assignment while having completely different learning experiences. One may have used AI to challenge their thinking, explain a difficult concept, test their understanding, or receive feedback. The other may have asked the AI to do most of the work and copied the result. Both technically “used AI,” but what happened educationally was very different, thus why how AI is being used matters.
AI Can Improve Learning Too
It is equally important that legitimate concerns about AI do not lead us into another overly simplistic narrative, namely that using generative AI automatically weakens learning. The evidence does not support that conclusion either. In fact, several recent studies and meta analyses have found that generative AI can improve learning outcomes under certain conditions, particularly when it is deliberately designed or used as a tutor, scaffold, feedback system, or instructional support rather than simply functioning as an answer generator.
For example, a 2025 randomized controlled trial published in Scientific Reports found that students using a carefully designed AI tutor learned more in less time than students participating in an active learning classroom lesson. Students using the AI tutor also reported greater engagement and motivation (2). Recent meta-analyses examining dozens of studies have likewise generally found positive average effects of generative AI on learning, while also finding substantial differences depending on the subject being taught, the educational level of the student, the duration of AI use, instructional design, and how the technology was actually incorporated into the learning process (3)(4).
For us, this is where the conversation becomes much more interesting than simply asking whether AI is “good” or “bad” for education. There is an enormous educational difference between a student asking AI, “Write my essay,” and asking, “Here is my argument. Challenge it and identify three weaknesses I should consider.” There is a significant difference between asking, “Give me the answer to this math problem,” and saying, “Don’t give me the answer. Give me one hint that will help me figure out the next step.” There is also a difference between asking AI, “Summarize this chapter so I don’t have to read it,” and saying, “I read this chapter. Ask me five questions that will test whether I actually understood it.”
In every one of these examples the student is using artificial intelligence, but the cognitive work being performed by the student is very different. This is why we believe the better educational question is not simply, “Did the student use AI?” but rather, “What role did the AI play in the student’s thinking and learning?”
The Temptation of the Cognitive Shortcut
One of the most important concerns surrounding AI in education has less to do with the technology itself and more to do with human behaviour. Learning can sometimes be uncomfortable. There are moments when a student has to wrestle with a difficult paragraph, struggle through a math problem, rewrite a weak sentence, retrieve information from memory, organize an argument, test an idea, make a mistake, reconsider an assumption, or sit with the frustration of not immediately knowing the answer. That mental effort is not necessarily evidence that learning has failed, sometimes that effort is part of how learning occurs.
Generative AI introduces something relatively new into that experience, that being an extraordinarily easy exit ramp from the difficult part. When the reading becomes complicated, AI can summarize it. When the essay becomes difficult, AI can write it. When the math problem becomes frustrating, AI can solve it. When the student cannot immediately think of an argument, AI can generate one. This can obviously be useful in some circumstances, but it also creates a temptation to reach for assistance at precisely the moment when the student’s own cognitive effort may be most educationally valuable.
Research led by Hamsa Bastani, a professor at the Wharton School of the University of Pennsylvania, provides one of the clearest illustrations of this problem. Bastani and her colleagues conducted a study involving nearly 1,000 high school mathematics students who were given access to different forms of GPT based assistance. Students using a general purpose GPT assistant performed significantly better while they had access to the AI. On the surface, this looked like an educational success. However, when the AI was subsequently removed and students were required to complete an examination independently, those who had used the unrestricted GPT system performed worse than students in the control group (5).
At first glance, AI had improved the student’s performance. However, once the AI disappeared, some of that apparent advantage disappeared with it. This is an important distinction because performance while using a tool is not necessarily the same thing as learning that remains after the tool is removed.
However, there was another part of the Bastani study that we believe is just as important and often gets overlooked. Another group of students used a specially designed GPT tutor that incorporated educational safeguards. Rather than simply providing answers, the system was designed to offer hints, encourage students through the problem solving process, and draw upon teacher developed solutions and common student errors. Those students received substantial assistance during practice without experiencing the same significant learning penalty when subsequently tested without AI.
This distinction matters enormously, because the study does not demonstrate that AI inherently harms learning. Instead, it suggests something considerably more useful, AI can improve immediate performance while still interfering with learning when it performs too much of the cognitive work for the student. However, when AI is deliberately designed and used as a learning scaffold rather than an answer machine, that risk can be reduced.
Performance Is Not Necessarily Mastery
This leads to another important distinction for parents, caregivers, and educators. AI assisted performance and independent mastery are not necessarily the same thing. A student may submit an excellent essay but struggle to discuss its central argument. A student may submit functioning computer code but be unable to explain why it works. A student may receive the correct answer to a math problem but not understand the process necessary to reproduce it independently. A student may generate an impressive research summary without ever reading the underlying research.
This does not mean AI caused the problem in every case, nor does it mean students who use AI are not learning. What it does mean is that the sophistication of a finished product may no longer tell us as much about the student’s independent knowledge and capabilities as it once did. A beautifully written essay can demonstrate that a high quality essay exists, but in an AI rich educational environment, it may not necessarily demonstrate that the student independently possesses all the skills reflected in that essay. We have also been hearing from educators who describe students doing well on projects and take home assignments but performing worse than expected when required to demonstrate the same knowledge independently without technology.
These are anecdotes, and we believe it is important to identify them as such. They do not establish a national trend, demonstrate causation, or substitute for peer reviewed research. However, teachers have historically been important early observers of changes occurring inside classrooms, much like the proverbial canary in the coal mine. What they are describing raises a legitimate research and educational question, “Could AI allow some students to maintain the appearance of academic competence while masking gaps in actual understanding?”
The Bastani research suggests that, under at least some conditions, the answer can be “yes”. However, that should not lead educators to automatically assume that polished student work was created by AI or that strong work should somehow become suspicious. Instead, it suggests that educators may increasingly need additional ways of making student thinking visible.
Assessment May Need to Change
This may become one of the biggest educational consequences of generative AI. Traditionally, teachers could often infer at least something about student knowledge from the work a student produced. Generative AI complicates that relationship because the sophistication of the final product may increasingly exceed the student’s independent mastery. This does not make the finished product meaningless, but it may mean that educators need additional evidence of learning.
Rather than assessing only the final product, educators may increasingly need to consider:
Can they explain how they produced it?
Can they defend the decisions they made?
Can they identify where AI contributed?
Can they verify the information?
Can they apply what they learned to a different problem?
Can they still demonstrate the underlying skill when AI is unavailable?
A teacher might ask a student to explain their reasoning, identify the evidence supporting their conclusion, describe what they disagreed with in the AI response, solve a related problem without AI, identify weaknesses in a generated answer, or explain what they changed and why. A student who used AI to help construct an essay might be asked to discuss the argument orally. A student who used AI while coding might be asked to explain several sections of the code or modify it to solve a slightly different problem. Suddenly, assessment becomes less focused on determining whether AI was used in the assignment, and more focused on determining whether the student actually understands what they submitted.
We believe this approach has another significant advantage. It shifts the educational conversation away from trying to “catch” students using AI and toward determining whether learning actually occurred. Rather than relying on AI detection software to determine whether AI may have contributed to an assignment, an approach that has repeatedly raised concerns about accuracy and reliability, educators can focus on something far more meaningful, that is asking students to demonstrate the knowledge, reasoning, understanding, and skills the assignment was actually intended to develop.
AI May Change Thinking Rather Than Simply Reduce It
There is another important nuance that we believe we should be alive too. Artificial intelligence does not necessarily eliminate critical thinking, sometimes it changes where the thinking occurs. Research involving adult knowledge workers has found that greater confidence in generative AI can be associated with lower self reported critical thinking effort. However, researchers also found that AI can shift cognitive work toward activities such as verifying information, integrating responses, supervising outputs, and deciding whether an answer is appropriate (6).
We need to be cautious about applying research involving adult workers directly to children and teens, but the finding raises an important educational question. If AI relocates cognitive effort, where is that effort being relocated to? If a student no longer spends twenty minutes struggling to locate basic information but instead spends that time comparing sources, challenging assumptions, verifying claims, and developing a stronger argument, that may represent a useful change in cognitive effort. If the student spends twenty seconds copying an AI generated answer and moves on without understanding it, that is something entirely different.
This is why statements such as “AI makes students think less” are probably too simplistic. In some circumstances AI may reduce cognitive effort, but in others it may redirect that effort toward different and potentially more sophisticated forms of thinking. The educational question is whether that redistribution supports the learning objective or simply allows the student to bypass it.
Not All Cognitive Friction Is Good
We also need to be careful not to idealize struggle, not all educational friction produces learning. Confusing instructions, inaccessible material, language barriers, missing prerequisite knowledge, learning differences, or the absence of timely feedback can interfere with learning rather than strengthen it. In these circumstances, AI may sometimes help remove barriers that prevent a student from engaging with the learning in the first place.
As one example, we spoke with a high school student who told us she was having difficulty understanding the way a particular subject was being taught. Rather than using AI to do the work for her, she would take the notes she had written during class and uses AI to reframe and explain the material in a way that made more sense to her. In this case, AI was not replacing the teacher or the student’s thinking, it was helping bridge a gap between how the material was being taught and how the student best understood it.
A student who does not understand an explanation might ask AI to explain the concept using simpler language or another example as the example above clearly demonstrates. A student learning English might use AI to help understand unfamiliar vocabulary. A student who benefits from information being presented differently may be able to ask an AI system to reframe or reorganize material. A student struggling with a concept while studying at home may receive immediate feedback rather than waiting until the next school day. These can all represent legitimate educational uses of AI and something we have seen students who have receive AI literacy education taking advantage of.
The challenge is distinguishing between unproductive friction that prevents learning and productive cognitive effort that develops knowledge, reasoning, memory, creativity, and problem solving ability. We should be cautious about using AI to remove the second while believing we are simply removing the first. Sometimes the struggle is not an obstacle standing in the way of learning, sometimes working through that struggle is part of how the learning takes place.
What About Student Voice?
We have also been hearing from educators who believe some submitted work is increasingly beginning to sound and look alike and this should not necessarily surprise us. If thirty students ask similar AI systems similar questions about the same assignment, the responses may differ in wording while still sharing similar structures, arguments, examples, vocabulary, and rhetorical patterns.
Emerging research is beginning to examine whether widespread AI assistance could contribute to greater homogenization of written expression while simultaneously improving fluency and efficiency (7). The evidence here is still developing, so we should be cautious about drawing broad conclusions. However, it raises an important educational issue worth watching, “Could widespread reliance on similar generative systems gradually narrow variation in student voice, argumentation, and expression?”
That questions matters educationally for several reasons. Developing a personal voice requires experimentation, awkward sentences, unusual connections, failed arguments, revision, disagreement, and eventually discovering how you communicate an idea. Those differences also give educators insight into how individual students are thinking. If AI increasingly mediates that process, a polished assignment may reveal less about the student’s reasoning than it once did.
There is also a broader educational concern. Classrooms benefit from intellectual diversity. Twenty students approaching the same question in twenty somewhat different ways can expose the class to different assumptions, experiences, arguments, metaphors, and solutions. If those students increasingly consult systems trained on overlapping datasets and optimized toward similar patterns of helpful, coherent responses, some of that variation could potentially narrow.
Learning to write is not simply learning to arrange grammatically correct sentences. Writing teaches students to organize thoughts, develop arguments, communicate uncertainty, discover what they believe, revise ideas, and gradually develop their own intellectual voice. AI can absolutely help with this process by providing feedback, identifying weaknesses, helping a student brainstorm, challenging an argument, or suggesting another perspective. However, there remains an important difference between AI supporting the author and AI becoming the author.
Blocking AI Does Not Teach AI Literacy
Given these concerns, some parents, caregivers, and educators may understandably want to pressure schools to block generative AI tools. There may absolutely be educational environments, assignments, examinations, developmental stages, and circumstances where restricting AI use is appropriate. Students still need opportunities to demonstrate what they can do independently, and as AI becomes increasingly capable, we would argue that protecting some of those AI free opportunities may actually become more important rather than less.
However, blocking AI should not be confused with teaching students how to use AI responsibly. A student who cannot access an AI chatbot on a school Chromebook may still have access through their smartphone five minutes after leaving school. More importantly, the larger digital environment is moving toward AI integration rather than away from it. Search engines are incorporating AI generated answers, word processors are adding AI writing tools, smartphones are integrating AI assistants, social media platforms are introducing AI characters and chatbots, and software used in workplaces is increasingly adding AI functionality.
Students will therefore increasingly encounter artificial intelligence whether schools formally introduce it or not. If our only educational response is to block the tool, we may prevent its use temporarily, but we have not necessarily taught the student what to do when they encounter it somewhere else. If our response is only to block a tool, we have not taught the child what to do when they eventually meet it somewhere else, and this is where AI literacy becomes essential.
What Does AI Literacy Actually Look Like?
In 2026, the OECD and European Commission released an AI Literacy Framework for primary and secondary education that describes AI literacy broadly as the knowledge, skills, and attitudes learners need to understand AI, critically evaluate it, and use it responsibly, ethically, and creatively (8). This broader definition is important because AI literacy should not simply become another term for knowing how to write a good prompt. For families and classrooms, we believe four practical questions should be asked and answered:
1/ Can a student explain what the AI gave them?
2/ Can they check whether the information is accurate and whether the sources actually support the claims being made?
3/ Can they exercise judgement about whether an output should be trusted, modified, challenged, or rejected?
4/ Can they recognize when AI is simply the wrong tool for the situation?
A student may produce a polished answer using AI and still be unable to explain where the information came from, whether the sources actually exist, what assumptions are embedded in the response, what important information might be missing, what might be incorrect, what evidence could change the conclusion, which parts represent their own thinking, how the AI influenced their thinking, or whether AI was appropriate for the task in the first place. They also need to learn something that may become increasingly important as these systems become more capable, when to put the AI down and bring another human being into the conversation. These are not simply classroom skills, they are increasingly becoming life skills.
Age and Development Matter
Another issue that sometimes disappears from discussions about AI in education is developmental readiness. A Grade 3 student, a Grade 8 student, and a Grade 12 student are not simply smaller or larger versions of the same learner. They are developing different foundational skills, knowledge bases, executive functions, metacognitive abilities, and capacities to evaluate information.
What constitutes helpful AI scaffolding for an older student may substitute for foundational learning in a younger student who is still developing literacy, numeracy, writing, memory, and reasoning skills. The research specifically examining generative AI in K-12 education is also still developing and is considerably less mature than much of the research involving university students and adults (9). This means schools should be cautious about taking research conducted with university students or adult professionals and automatically assuming that the same findings apply to a ten year old child.
We do not believe schools should approach AI with one universal rule that applies equally across every grade and every subject. The question should not simply be, “Should students use AI?”, it should also be, “At this developmental stage, for this student, with this learning objective, what role should AI play, if any?”
There May Be an Emerging AI Literacy Divide
For years, discussions about technology and education have focused on the “digital divide,” particularly unequal access to devices and internet connectivity. Artificial intelligence may introduce another kind of divide, and it may ultimately prove just as important. The meaningful difference may not simply be between students who have access to AI and students who do not, it may increasingly be between students who have been taught how to use AI well and students who have not. This is something we have observed firsthand through our work presenting to a wide range of schools and engaging directly with both educators and students.
PISA 2025 found socioeconomic differences in student’s reported opportunities to learn how to evaluate AI generated information (1). This matters because a student surrounded by parents and educators who teach them to question AI, verify claims, protect personal information, recognize limitations, use appropriate prompts, understand bias, and know when to work independently may interact with these systems very differently from a student who simply discovers an AI chatbot and begins using it as an answer machine.
If we are not thoughtful in our approach to AI in schools, we may create the beginnings of an AI literacy divide, something that Erin Mote spoke so eloquently about in a recent Forbes Article (12). Erin’s thoughts in the article, as well as a posting on her LinkedIn page, “If we push AI out of the classroom, we leave a massive gap. Students with high speed home broadband and personal devices will keep building AI literacy. However, students relying on metered, fragile connections will be shut out entirely, losing out on guided, safe learning experiences and future economic opportunities.”
We completely agree with Erin’s thoughts on this issue. The educational advantage may eventually belong not simply to students who have access to the most powerful artificial intelligence, but to students who have parents, caregivers, and educators around them teaching them when to use it, how to question it, how to verify it, how to challenge it, and when to put it down. This is another reason why AI literacy should not become an educational opportunity available primarily to students who already have advantages.
This is something we have also observed anecdotally through our work with both independent and public schools. In our experience, staff and students in independent schools have often demonstrated a greater level of familiarity with the pedagogical use of AI, including how these tools can be thoughtfully integrated into teaching and learning. This is not to suggest that this is true of every independent or public school, but it does raise an important question about whether differences in access, professional development, resources, and institutional flexibility could contribute to an emerging AI literacy gap between educational settings.
While some families and schools can afford premium AI tools that offer more advanced capabilities, higher usage limits, and access to newer models, other families may have to rely on free or more limited versions, or may have little access at all. This creates the potential for a new kind of educational divide, where some students have greater access to powerful AI tools that can support research, tutoring, feedback, creativity, and problem-solving, while others are working with fewer or less capable resources.
Yes, Sometimes Students Should Work Without AI
Teaching AI literacy does not mean every assignment needs AI. In fact, we believe the opposite may be true. As AI becomes increasingly available, schools may need to become more deliberate about protecting opportunities for students to think without it. There is educational value in asking students to write something themselves, do mental arithmetic, retrieve information from memory, read an entire chapter rather than asking for a summary, struggle with a difficult question, participate in a classroom discussion where nobody can instantly outsource the answer, and experience being wrong before figuring out why.
There is also value in learning when AI genuinely improves the learning process. This is why the goal should not be to eliminate cognitive effort from education simply because technology makes that possible. Before allowing AI for an assignment, an educator might ask a simple question, “What cognitive work am I trying to develop here, and will AI support that work or perform it for the student?” We believe that may become one of the most important instructional questions when it comes to integrating AI into the classroom.
Parents and Caregivers Have a Role Too
This conversation should not stop at the classroom door. Parents and caregivers do not need to become artificial intelligence experts, but we do believe families should begin talking about the difference between using AI to learn and using AI to avoid learning. Instead of only asking, “Did you finish your homework?” parents and caregivers might occasionally ask,
Tell me what you learned?
What part was difficult?
Did you use AI?
What did it help you understand?
What thinking did you do yourself?
Was there anything the AI got wrong?
How did you check it?
Could you explain this without the AI?
These questions are not about catching a child doing something wrong, they are about making the child’s thinking visible and helping them develop the habit of reflecting on how technology is influencing the way they learn. Ultimately, one of the most valuable AI literacy skills we can give young people may be the ability to ask themselves, “Is this tool helping me think, or is it doing the thinking for me?”
At The White Hatter, we do not believe the educational conversation surrounding artificial intelligence is well served by framing AI as either a miracle for education or a catastrophe that must simply be banned, both of these positions are too simplistic. Generative AI can potentially help students brainstorm, receive explanations at different levels, practise concepts, explore ideas, receive immediate feedback, overcome some accessibility barriers, translate information, challenge arguments, and support creativity. Under well designed educational pedagogical conditions, research suggests it can improve learning.
However, the same technology can hallucinate information, reproduce bias, encourage over reliance, weaken source evaluation, facilitate academic dishonesty, flatten individual voice, and allow students to bypass some of the cognitive work necessary for learning. Both realities can exist at the same time, which is why we believe AI literacy is becoming as important as AI access.
Our youth and teens are growing up in a world where increasingly sophisticated artificial intelligence will be available almost everywhere. The challenge for education is therefore not simply teaching young people how to get answers from these systems, it’s teaching students what to do with those answers:
Question them
Verify them
Challenge them
Improve them
Compare them with other evidence
Know when to use them
Know when not to use them, and
Understand what thinking still belongs to the student, and remain capable of thinking when the AI is not there.
Perhaps the educational outcome that matters most in the age of artificial intelligence is not whether a student can successfully use AI, it’s whether they can use AI without surrendering the thinking, judgement, curiosity, creativity, and intellectual independence that education is supposed to help them develop.
The educational challenge presented by generative AI is therefore not simply preventing students from using it, nor is it encouraging students to use it simply because the technology exists. It is preserving the cognitive work that learning requires while using AI where it can genuinely support that work. That is considerably more difficult than simply blocking an app, however, we believe it is also a much better way to prepare students for the onlife world they are entering where AI is being integrated into everything.
There is little doubt that generative AI is creating a significant disruption to many traditional educational and pedagogical practices. However, the disruption is not simply about students having access to a new tool that can write an essay, answer a question, or complete an assignment. AI is challenging some of the assumptions that have shaped education for decades, including how we teach, how students learn, how assignments are designed, how knowledge and understanding are assessed, and ultimately what skills young people will need when they leave school.
This is why the conversation cannot be reduced to whether schools should allow, restrict, or block AI. Parents, caregivers, educators, students, and educational leaders need to be having much deeper conversations about where AI can genuinely support learning, where its use may interfere with the development of foundational knowledge and skills, and where students need opportunities to think, struggle, create, reason, and problem solve without AI assistance. Just as importantly, we need to consider whether our traditional methods of assessment can still reliably tell us what a student actually knows and can do when powerful generative tools are readily available.
Preparing students for this educational disruption therefore requires more than teaching them how to prompt an AI system. It requires AI literacy, critical thinking, verification skills, intellectual independence, ethical decision making, and an understanding of when using AI strengthens learning and when it may replace the very cognitive work that learning requires. The goal should not be to prepare students for a world without AI, because that world no longer exists. The goal should be to prepare them to enter an AI integrated world with the knowledge, skills, judgment, and human agency (13) to use these technologies as tools for thinking rather than substitutes for it.
Digital Food For Thought
The White Hatter
Facts Not Fear, Facts Not Emotions, Enlighten Not Frighten, Know Tech Not No Tech
References
