AI in Schools? - The Better Question Isn’t Whether We Should Use It, But When, Why, and How
- The White Hatter
- 6 minutes ago
- 11 min read
Caveat - Over the past year, we have seen a significant and, we believe, necessary increase in conversations surrounding the role of artificial intelligence in school curricula. With a new school year just around the corner, we expect these discussions among parents, caregivers, educators, administrators, and school districts will only intensify. This article is our contribution to that evolving conversation, with the hope of adding some important context, nuance, and perspective to discussions that can too easily become framed as simply being “for” or “against” AI in education.
A debate unfolding in school districts across North America raises a question that we believe schools, parents, caregivers, and educators will increasingly need to consider, “Should parents be required to provide permission before artificial intelligence is used as part of their child’s education?”
As an example, according to a proposal being discussed in Florida, guidelines would include parental notification when an AI system has been approved for use and would give parents the opportunity to opt their child in when a class is going to use it (1). Supporters argue that this provides families with greater control over their child’s exposure to a rapidly evolving technology. Others argue that an opt-out approach would be more practical, particularly given that permission forms and school communications are frequently missed or never returned.
We understand why some parents and caregivers would find an opt-in approach appealing. Artificial intelligence is developing incredibly quickly, there are legitimate questions surrounding privacy and data collection, and the long term effects of some forms of AI use in education are still being studied. Schools have not always done a particularly good job introducing new educational technologies either. Too often, technology has been purchased first and questions about privacy, pedagogy, safety, teacher training, and educational value have been asked later.
However, we think there is a fundamental problem with framing this debate simply as, “Should parents have to approve AI?”, given that AI is not one technology, used in one way, presenting one level of risk.
The more important questions should be, “What is the AI actually doing?”, “Who is interacting with it?”, “What information is being provided to it?”, “Why is it being used?”, “What evidence supports that educational purpose?”, “What happens if the technology gets something wrong?” and “What safeguards are in place?” The answers to those questions, rather than the presence of the letters “AI,” should help determine what level of parental notification, choice, or consent is appropriate.
One argument we sometimes hear is that parents and caregivers should simply be allowed to decide whether they want AI used in their child’s classroom. Parental involvement absolutely matters, and parents and caregivers should have meaningful information about technologies being used with their children in school. However, informed consent requires something more than simply sending home a permission form containing the words “artificial intelligence.”
Most parents, caregivers, and even educators are not AI specialists, nor should we expect them to be. Many are still trying to understand the difference between traditional algorithms, generative AI, large language models, AI-powered search, adaptive learning systems, automated assessment, and conversational chatbots. They are also being exposed to a constant stream of conflicting messages about AI, ranging from claims that it will revolutionize education to warnings that it will destroy critical thinking.
That doesn’t mean parents and caregivers are incapable of making decisions about their children. It means schools have a responsibility to provide parents and caregivers with enough understandable information to make those decisions meaningfully.
Parents and caregivers shouldn’t have to become AI experts. Schools should be able to explain, in plain language, what an AI system does, why educators want to use it, what educational objective it supports, what student information it receives, whether that information leaves the school environment, how long information is retained, whether it is used to train an AI system, what the risks are, and what safeguards have been established.
However, we should apply the same standard to educators and school administrators. Working in education does not automatically make someone knowledgeable about artificial intelligence. Teachers, administrators, technology departments, and school districts also need appropriate AI literacy and professional development. The answer to limited parental understanding cannot be blind institutional trust. It should be informed institutional decision making combined with meaningful parental transparency.
This is where we believe much of the current conversation about AI in education becomes unnecessarily polarized. Consider a teacher who uses an approved AI system to generate ten fictional math problems for tomorrow’s lesson. Now compare that with a school uploading a student’s psychological assessment, behavioural history, academic records, or personally identifiable information into an AI system. Both technically involve “using AI”, however, from a privacy, safety, ethical, and educational perspective, they are clearly not equivalent.
The same applies to a teacher using AI to generate discussion questions, create differentiated reading examples, translate teacher created classroom instructions, brainstorm hypothetical scenarios, or demonstrate how an AI chatbot can confidently produce incorrect information so students can practise fact checking.
Compare those uses with an AI system evaluating a student’s academic performance, profiling their behaviour, recommending whether they should receive additional supports, determining their educational pathway, assessing whether they present a behavioural risk, or influencing disciplinary decisions.
Again, all could be described as “AI in education,” but their potential consequences are dramatically different. This is why we believe permission should be based primarily upon consequence and risk, not simply the presence of AI.
There is another distinction that we believe deserves far more attention. Learning about AI is not necessarily the same as learning with AI, and neither is the same as being assessed by AI.
Learning about AI could involve teaching students how generative AI works, why AI systems hallucinate, how bias can appear in outputs, how deepfakes are created, how AI generated misinformation spreads, what happens to information entered into AI systems, why AI answers require verification, how synthetic content can influence us, and when AI should not be used.
Learning with AI involves students actually using AI as part of the learning process. They might use it to brainstorm ideas, practise another language, receive tutoring assistance, explore alternative explanations of a difficult concept, generate practice questions, assist with coding, or critique an AI-generated answer.
Being assessed by AI is something different again. An AI system may be grading, profiling, predicting, monitoring, categorizing, recommending, or otherwise influencing decisions about a student.
These three activities should not automatically be governed by exactly the same rules. A parent might reasonably object to their 12-year-old having unrestricted conversations with a generative AI chatbot while strongly supporting their child being taught how generative AI works, how it can manipulate information, and how to recognize when an AI-generated answer is wrong, and that is an important distinction.
Rather than creating one universal rule requiring parental permission whenever AI appears somewhere in the educational process, we believe schools should consider a graduated, risk based approach.
At the lowest level might be teacher facing AI, where an educator uses an approved system for lesson planning, brainstorming, creating fictional examples, developing practice questions, or preparing instructional materials without entering identifiable student information. Requiring individual parental permission for every such use would be difficult to justify.
The next level might involve student facing instructional AI, where students interact directly with an approved AI system for a defined educational activity under teacher supervision. This deserves greater transparency, clear educational objectives, age appropriate safeguards, and potentially parental notification or choice depending upon the technology and circumstances.
A higher level would involve student data AI, where identifiable student work, learning information, behavioural information, or other personal data is processed by an AI system. Here, privacy assessments, data minimization, security requirements, vendor scrutiny, transparency, and meaningful parental involvement become considerably more important.
At the highest level would be consequential AI, where artificial intelligence grades, profiles, predicts, recommends placement, assesses risk, influences discipline, evaluates behaviour, or otherwise materially affects a student’s educational experience or opportunities. These applications should receive the greatest scrutiny, meaningful human oversight, and potentially explicit parental consent or even restrictions depending upon the application.
This approach recognizes something that blanket AI policies often miss, that being risk comes from what the technology is doing, not simply what the technology is called.
We also think it is important that those advocating for AI literacy acknowledge that concerns about artificial intelligence in schools are not simply the product of parents being frightened by new technology. We acknowledge and often speak to legitimate challenges in our written work and in our programs for students, parents, and educators.
Generative AI systems can produce false information with extraordinary confidence. They can reproduce bias contained within their training data. Some services collect information about users and their interactions. Students may inadvertently disclose personal or sensitive information. AI generated material can contain inappropriate or misleading content. Questions remain surrounding copyright and intellectual property. Excessive dependence upon AI could also encourage cognitive offloading, where students increasingly delegate thinking, writing, problem-solving, or creativity to machines rather than developing those abilities themselves. These are legitimate educational concerns and should not be dismissed as technophobia, in fact, they are precisely why schools need thoughtful AI governance.
However, recognizing that AI presents risks does not logically mean that every educational use of every technology containing artificial intelligence should require individual parental authorization. The presence of risk calls for proportionate safeguards, not necessarily universal prohibition or universal permission slips.
There is another question that should also be asked before any AI system enters a classroom, “What educational problem are we trying to solve?” Using AI simply because it is new, impressive, or because a technology company has successfully marketed it to a school district is not a pedagogical justification. AI should not be introduced into a classroom and then have educators search for reasons to use it, the educational objective should come first.
If a teacher believes AI could help students understand a difficult concept, practise critical evaluation, explore competing perspectives, receive differentiated support, or develop AI literacy, then the next question should be whether there is evidence and sound pedagogical reasoning supporting that particular application. Sometimes the answer may be yes, and sometimes the answer may be no. Sometimes the best educational technology available to a teacher will still be a pencil, a piece of paper, a book, a conversation, or another human being.
Responsible AI integration should never begin with the assumption that AI improves learning simply because AI is being used. This is particularly important when AI begins influencing decisions about students. Artificial intelligence can assist educators, but it should not replace professional teaching judgment or meaningful human decision making.
A teacher might use AI to suggest possible feedback on a student’s writing, but the teacher should remain responsible for determining whether that feedback is accurate and educationally appropriate. An AI system might identify patterns in student performance, but educators should understand the limitations of those predictions before making decisions based upon them.
The greater the consequence for the student, the stronger the requirement should be for meaningful human review. Put another way, “AI can inform educational judgment, however, it should not become educational judgment.
There is another unintended consequence when we make responsible, educator guided AI use in the classroom unnecessarily difficult. Restricting AI at school does not mean students will stop using it. Many young people already have access to powerful AI tools through their phones, home computers, search engines, social media platforms, productivity software, gaming environments, and a growing number of everyday technologies. If schools simply restrict or avoid AI rather than teaching students how to use it critically, ethically, safely, and responsibly, much of that learning will take place outside the classroom, often without the guidance of a knowledgeable adult.
We also need to consider where young people actually spend their time. Schools may have a student’s attention for six or seven hours a day, five days a week, and sometimes longer when sports, clubs, or other extracurricular activities are included. That still leaves a significant portion of the day outside the school environment. This is why the conversation cannot be framed as a choice between schools or parents controlling a child’s exposure to AI. Young people are already growing up in a world where AI is increasingly embedded into the technologies they use. The better question is how parents, caregivers, and educators can work together to ensure that young people learn how to navigate these technologies with the critical thinking, judgment, ethics, and digital literacy they will need when trusted adults are not standing beside them.
If AI disappears from supervised educational environments while remaining readily available everywhere else, we risk creating what we have previously described as a “digital blind spot.” However, the argument for AI education should not simply be, “Kids are going to use it anyway.” The stronger argument is that AI literacy itself is becoming part of modern literacy.
Young people increasingly need to understand how AI generated information is produced, why AI can be wrong, how synthetic media can manipulate perception, how to verify AI generated claims, what happens to information they provide to AI systems, how AI can reinforce bias, where AI may legitimately assist them, and when using AI may actually undermine their own learning. Those are educational competencies. We believe that schools are uniquely positioned to help young people develop them with knowledgeable adults present.
None of this diminishes the role of parents and caregivers. Parents deserve meaningful transparency about technologies being used with their children. They deserve understandable explanations rather than pages of technical language. They deserve to know when their child’s information is being collected, analyzed, retained, transferred, or potentially used for purposes beyond the immediate classroom activity.
There will also be circumstances where explicit parental consent is entirely appropriate, particularly when children are creating individual accounts with external AI providers, identifiable information is being processed, sensitive information is involved, persistent student profiles are being created, or automated systems are materially influencing decisions about a child.
However, parental involvement should mean more than receiving another electronic permission form buried among dozens of school communications. If schools genuinely want informed parental participation, they need to help families understand the technology first. Consent without understanding isn’t necessarily informed consent.
Perhaps the biggest mistake we can make is turning artificial intelligence in education into another binary debate. You either support AI in classrooms or you want it banned. You either believe AI will transform education or you believe it will destroy it. Neither position leaves much room for thoughtful educational decision making.
We believe a better approach is evidence informed, pedagogically justified, transparent, human centred, and proportionate to risk. Before approving an AI application, schools should be asking:
What educational purpose does this serve?
Is there evidence or sound pedagogical reasoning supporting that purpose?
Is the student interacting directly with the AI?
What information is being entered into the system?
Is personally identifiable or sensitive information involved?
Where does that information go?
How long is it retained?
Can it be used to train or improve the AI system?
What happens when the AI gets something wrong?
Is a qualified human reviewing consequential outputs?
What alternatives exist for students and families?
What level of parental notification, choice, or consent is proportionate to the actual risk?
Those questions tell us considerably more than simply asking, “Does this use AI?”
The Florida discussion raises an important issue, and we believe parents should absolutely be part of that conversation. However, requiring parents and caregivers to opt their children into every educational use involving artificial intelligence risks treating dramatically different technologies and activities as though they present exactly the same concerns, and as discussed in this article, they don’t.
A teacher asking an approved AI system to create fictional math problems is not equivalent to uploading a student’s psychological assessment.
Teaching students how to identify an AI hallucination is not equivalent to allowing a child unrestricted access to an AI companion.
Using AI to translate teacher created classroom instructions is not equivalent to allowing an algorithm to decide whether a student belongs in an advanced academic program.
Calling all of these activities “using AI” tells us very little about their actual educational value or level of risk. Perhaps the better principle is this, the presence of AI should not, by itself, determine the level of parental consent required. The purpose, data, interaction, risk, and potential consequence of its use should.
Parental consent remains an important safeguard, but a permission slip should never become a substitute for evidence, governance, privacy protection, professional competence, transparency, and sound pedagogy.
The challenge facing schools isn’t simply deciding whether children should be allowed to use AI. Rather, it’s helping young people learn when AI is useful, when it isn’t, when it can be trusted, when it needs to be questioned, when it needs to be verified, and perhaps most importantly, when they are better off thinking and doing the work themselves. That is not teaching children to depend upon artificial intelligence, it’s teaching them how not to.
Digital Food For Thought
The White Hatter
Facts Not Fear, Facts Not Emotions, Enlighten Not Frighten, Know Tech Not No Tech
Reference:
