Students Are Using AI, Now What?
Navigating the AI Dilemma in Schools: Sanctioned, Unsanctioned, and Everything in Between
When it comes to AI use by students, I want to distinguish between three very different kinds of use. I'll go over each to explain my views because it's important we don't lose sight of what matters for the children. While some use cases, as we'll see, can be useless at best and counterproductive at worst, we should not throw the baby out with the bathwater either.
Incognito Mode AI
First, we have the sneaky unsanctioned use of AI. That's when students use AI behind your back and pass generated text and ideas as their own. Depending on which survey we're reading, we can probably agree that about half of all secondary students are aware of the technology and have used it at least a few times.
The problem with this kind of use is twofold. On one hand, students are outsourcing their cognitive work to AI, which means they’re not building the skills needed to assess the AI’s output or honing their critical thinking and communication skills. Essentially, they’re wasting their own time and that of the teacher who ends up grading AI-written papers, which can also lead to conflict. On the other hand, because the use of AI is unsupervised, students miss out on feedback regarding their AI literacy skills. Could they have used this tool to produce better work? Probably, but they’ll never know because they’re keeping it a secret. In short, they’re not developing the intended skills from the assignment, nor are they advancing their AI abilities.
This type of usage will always occur, but steps can be taken to mitigate its occurrence. For instance, avoid assigning work that relies on students not using AI, or rethink the design of assignments altogether—more on that later.
The Shiny-New-Plastic-Toy AI
The second kind of AI use by students is the sanctioned use of AI gadgets like personalized tutor from Khanmigo or SchoolAI bots. This is likely to be the fastest-growing trend as more "AI" companies (read my previous post on AI tools to understand the quotation marks) roll out student-facing apps promising to personalize learning by adapting to each child's learning styles, disposition, interests, maybe even zodiac sign, why not? Dan Meyer, one of the most eloquent critics of this new fad, brilliantly exposes the irony of these promises, which have been recycled time and time again by software companies for 30 years. Will this time be different? I doubt it as well. All of these tools boil down to students interacting with chatbots. And no matter how good the chatbot is or becomes, it doesn’t matter unless students actually engage with it. MagicSchool, for instance, recently released its student-facing platform offering 50 AI tools. Here are a few of them:
If you look closely, you'll quickly notice they're all the same: drop-down menus and small text boxes that send an API request to ChatGPT, then present the text response. In other words, students use ChatGPT without knowing it, and more importantly, without learning prompt engineering—currently one of the few economically valuable AI-related skills that can easily be taught to kids.
Furthermore, while conversing with a chatbot impersonating Frederick Douglass can feel almost magical to educators who remember the pre-internet world, it won’t have the same effect on 13-year-olds for whom this kind of technology will feel like old news by Thanksgiving. I've been thinking and saying so ever since the first student-facing "tools" started to come out, and therefore I was delighted to see Dan Meyer's recent Chatbot Tutor Benchmark that he graciously allowed me to share with you here.
Creating customized chatbots is simple, which is why they’re popping up everywhere. Most rely on ChatGPT or other foundation models, so why not teach students to build one themselves? To sum up, my beef with these new student-facing AI tools is that they amount to using new tools to do old work, and they don't do anything to advance AI literacy and proficiency. Another major flaw of these student-facing chatbots is that they're very easily jailbroken. “Jailbreaking” an AI refers to getting it to do something it's not supposed to do, such as revealing some of its training data (spying), designing chemical or biological weapons (terrorism), or simply spewing inappropriate language. While AI labs are working hard to prevent their chatbots from helping you design a bomb, they're understandably not that concerned with preventing Frederick Douglass from helping you cheat on your math assignment.
Before moving on, I have to admit that these student-facing chatbots present at least two advantages, one minor and one major. For a student who's completely new to AI chatbots (about half of students at the time of writing), the grid of 50 AI tools from MagicSchool can actually help them understand some of the technology's possibilities. Of course, students should be taught directly that under the hood, these tools are but one unique tool, and that they can easily learn to use this tool without the mediation of MagicSchool if they learn the basics of prompt engineering. This advantage, however, ceases to be one as soon as students get even a basic understanding of what they're working with.
Side note: this is exactly the same for almost all the teacher-facing tools offered by the various platforms out there, with the exception of tools that streamline or automate aspects of content delivery, grading, feedback, data analysis, etc., but that's LMS software engineering, not AI per se (I may write about this too).
Their major selling point is oversight. Not only do teachers have access to student conversations with the chatbots, which in itself, even when students are using the tools as intended, is extremely valuable since it provides a window into the students' thinking and AI skills, but all these tools also come with automated monitoring of the conversations and can raise red flags in case students are showing signs of distress or malaise, or are otherwise making good ol' Frederick Douglass do their homework for them.
"Wess - The student asked for help solving a math problem (x^2-9=0) instead of engaging with the historical conversation." SchoolAI bot reporting on my misuse
So far, not so good... We went over under-the-radar unsanctioned use of AI that needs to be addressed because outsourcing cognitive work without possessing the cognitive skills to assess the AI's output is a waste of time for everybody, and teachers are missing an opportunity to teach valuable AI skills to students. We also went over arguably useless AI gadgets that software companies promise are the best thing since sliced bread and will, this time, actually provide competent, compassionate, and personalized 1-on-1 tutors to every kid in the solar system.
Note that if they're correct, my job here is done, education is solved, we can go worry about something else now. If not, I invite you to keep on reading, because there's a third approach that holds immense potential. This method combines the best of both worlds: leveraging AI's capabilities while fostering critical thinking and AI literacy.
The Holy Grail AI
Get a load of this: sanctioned use of AI that fosters the development of key cognitive skills and content knowledge AND helps students develop their understanding of AI and how to get the best out of it.
"The world is changing and if we don't prepare our students to work with AI resources, we are not preparing them for the world. If we spend our time on making our courses AI-proof — assigning hand-written papers or such — we are missing the point of education. We need to teach our students how to have the AI think with them, not for them. This is the most important goal: if we don't achieve that, the AI will become their competitor."
Dr. Boris Steipe, Professor Emeritus, Biochemistry, University of Toronto
I can’t recall how this quote came to me, but it’s been a staple in my AI workshops. As educators, we cannot forget this fact: AI is not an EdTech, it's not a tutor, it's not this annoying tech thing that is forcing us to change the way we teach... AI is a general-purpose technology that will permeate every aspect of our lives and profoundly transform our world, starting with education, yes, but more importantly, the job market.
Do you remember 1995? Imagine falling into a coma in 1995 and waking up today, asking someone what smartphones and the internet are for. We're there now, just like in the late nineties: Amazon and Google are brand new, broadband internet is spreading, we don't know yet about smartphones and social media and influencers and YouTube, etc., but what is becoming abundantly clear is that this general-purpose technology called the internet will chart the course of the economy and will probably shape our lives. Teenagers and young adults of today are utterly incapable of performing the thought experiment I just proposed. They simply cannot fathom what the world was like prior to cell phones and broadband internet. Will it take 30 years for AI to transform the way we live beyond recognition to the point that future teenagers and young adults won't be able to fathom what the world was like in the early 2020s? I am convinced it will be much quicker than that; in fact, I'll take a chance and claim that only the oldest of Gen. Alpha, today in grades 3 to 6, will be able to vaguely recall what the world was like when you couldn't talk to your TV or computer.
So, how do we teach what we need to teach, help students understand AI, and use it to enhance their learning? This has been my quest for the last year and a half, and I believe every educator should be thinking about this when designing assignments. The goal is to leverage AI to help students master learning standards and develop AI skills.
Corporate leaders are already saying they’d hire someone with AI skills over someone more experienced but without AI proficiency1. And yes, AI can even help plumbers or landscapers—just ask ChatGPT, "How can AI like ChatGPT help a plumber in their business?"
I do have a few ideas—some I’ve even tried successfully with students of various ages—but I plan to experiment a lot more this school year. Hopefully, in a few months, I’ll have a top 5 or top 10 project list to try with your students. If you’re interested in experimenting, feel free to reach out to me. Aside from direct classroom work, I’m part of two national collaborative initiatives2 and trying to launch a similar initiative locally through Ulster BOCES. The goal is to build a community of teachers willing to design innovative AI learning experiences, test them, and share feedback. Interested? Please step forward!
For concrete ideas on how to integrate AI in your classroom this way, I'm afraid you can't rely on state-issued guidelines, 22 at this point (New York, Hello?). I'll just share here the document from the Washington State Education Department, which is by far one of the best out there and the only one to dedicate space to speak to students directly. No, the solution(s) will come from the ground up, not from an AI lab or a start-up or government office, but straight from our classrooms.
To conclude and leave you with a little nugget, I’ll share an upcoming project I have. In a previous article, I shared an idea for turning ChatGPT into a cover letter writing coach through advanced prompting. The goal was to show students how they could transform a chatbot into a highly useful, specific tool using a relatable example that would engage them and (hopefully) spark that all-important question: “What else can I do with this?” This sense of inquiry is exactly what I aim to cultivate in all my AI-related assignments and activities. Next week, I’ll be using the cover letter coach prompt with seniors. We’ll break it down, analyze how it works, and then they’ll be tasked with creating their own College Essay Writing Coach. To succeed, they’ll need to understand what makes a good college essay based on individual personalities and aspirations—and they’ll get hands-on experience with some of the most advanced prompt engineering I know. Wish us luck!
2024 Work Trend Index Annual Report, from Microsoft and LinkedIn
AI for Innovative Schools, from the Digital Learning Collaborative
The School Teams AI Collaborative, from Leading Educators and The Learning Accelerator