In my latest article and podcast, I share a continuum for how we can think about AI integration in our educational institutions.

Listen to the Podcast

If you enjoy this blog but you’d like to listen to it on the go, just click on the audio below or subscribe via iTunes/Apple Podcasts (ideal for iOS users) or Spotify.

 

How Should We Approach AI in Schools?

When ChatGPT first became publicly available, I had a conversation with my friend and collaborator A.J. Juliani (who, by the way, just wrote a phenomenal article on academic honesty and AI) about how schools might approach AI. I talked about the need to have an integrated approach that avoids the opposing pitfalls of Lock It and Block It and Techno-Futurism.

But we also began to reconsider what this might mean for our set of Boost PBL projects. Should we incorporate some kind of AI chatbot into the student interface? Should we design a student-facing AI tool that functioned as a project manager? We’re still playing around with ideas. However, we ultimately decided on making the projects human-centered and somewhat AI resistant.

In other words, the default is mostly low-tech. Students collaborate in person and talk through their ideas face-to-face. They have paper and pencil Boost Books. They do rotating reading with physical articles that they can access. In some cases, they have hands-on research as well. They sketch note ideas. They write ideas on sticky notes. They prototype with upcycled materials.

And yet . . .

We also have opportunities for tech integration. I am adding ideas of Boost Extensions that incorporate machine learning into the projects. This way, teachers are the ones who decide when and how they want to integrate AI into the projects in a way that is more intentional and rooted in their own knowledge of their students.

I share this story because so many schools are wrestling with the question, “How should we approach AI in teaching and learning?” And today, I’d like to share a continuum I developed as a tool for thinking about this question.

Four Approaches for Approaching Generative AI in Education

I developed the following model as a way to think about how we want to approach using generative AI for teaching and learning. I want to recognize from the get-go that others have developed great models for tech integration decades ago. For example, the SAMR model, which is a hierarchical model that moves progressively higher from Substitution and Augmentation (both used to enhance the learning) and Modification and Redefinition (both used to transform the learning).

What I want to share is a little different. It’s a sliding scale rather than a hierarchy. It recognizes the need for tech-resistance as well. While it can be used to think about single tasks or whole assignments, it’s also something we can use as we consider broader school-wide initiatives.

 

1. The AI-Resistant Approach

At the far side of the continuum, we have the AI-Resistant Approach, where educators actively resist AI assistance. It might come from a place of skepticism, where teachers feel like the power of AI has been overblown and overhyped. I can’t blame them. I’ve lived through the buzz of one-on-one devices, interactive whiteboards, and virtual reality. Over twenty years ago, Larry Cuban wrote a brilliant book Oversold and Underused about why the personal computer had failed to live up to its promise as a transformative learning device. The promise of instant information leading to better learning outcomes failed miserably and many teachers avoided educational technology altogether. The same might be true of machine learning.

For some teachers in this camp, the issue isn’t “I’m not sold on AI” so much as it is a fear of AI. Students will use AI to teach and academic dishonesty will be rampant. Those who use it will still fall victim to cognitive atrophy and grow too reliant on machine learning to do the thinking for them (an idea that was highlighted in research where students used AI chatbots in math). Others still are more concerned about a larger global trend of AI in our world. They’re worried about using tools in their classroom that might someday lead to the loss of jobs. They worry about AI’s environmental impact, in terms of both energy and water usage, on the environment.

It’s easy for technology enthusiasts to write this first level off as fear-mongering but their criticism is rarely cynical. Many of these teachers want to incorporate hands-on learning activities that embrace the tactile. They have seen the danger of too much unsupervised screen time and want to design learning experiences that emphasize human connection instead. They are rightfully worried about unintended consequences of a tool that we do not understand entirely.

What does this look like?

In a PBL unit, an AI-resistant approach would focus on hands-on learning and human interaction. The teacher would begin the unit with a concept attainment lesson, front-loaded vocabulary, and a roadmap for the project. This would help build background knowledge and reduce cognitive overload. From there, they would move into an entry event that might involve a field trip, an interview with an expert, a short read aloud, or anything else that gets students excited about the project. Students would then engage in inquiry and research. They would write their questions in a Boost Book, on sticky notes, or on chart paper. Research might involve a rotating reading activity, an interview with an expert, observational research, play-based research, experiments, or surveys. As they begin planning and prototyping, teachers would lean into low-tech tools and upcycled materials. Notice that this is AI-resistant but it’s not heavy handed or compliance-based. Instead, it is fully centered on the tactile, tangible, and human elements in the room.

For a class discussion, an AI-resistant approach might be a Socratic Seminar activity where students read a short text, annotate elements of the physical copy, and then engage in an open-ended dialogue:

For student writing, an AI-resistant approach would begin by crafting a writing prompt that leans into elements that humans do well. So, instead of asking, “What was the theme of The Great Gatsby?” a teacher would consider the role of curiosity, empathy, context, and voice. The new prompt would be, “What is a theme of The Great Gatsby and how does it play out in our school. Reference one thing someone shared in our Socratic Seminar and connect that to one of the lines in the book.” Or perhaps, “Who is the most unlikeable character in the story? Create a conversation between you and the character where you slowly gain empathy toward that character.” This AI-resistant approach might still include technology but it might also have sketchnotes, concept maps, and webs. It might involve a quick handwritten writing piece and a peer discussion.

For a choice-based activity, an AI-resistant approach might be a set of stations that students go to in a style similar to a Montessori School. It might be a Tic-Tac-Toe choice board activity with handouts. But it might simply be silent reading, where they get to read whatever book they checked out at the library. In math, it might be a “What can you do with this?” problem popularized by Dan Meyer, where students are given a scenario (often a video or picture) and develop their own math problem.

 

2. The AI-Assisted Approach

In some cases, teachers might see some value in using machine learning but only as a means of saving time and making tasks more feasible. In most cases, this AI-Assisted Approach keeps strict boundaries around using generative AI tools for direct teaching and learning but allows for AI to function on the back-end by creating rubrics, designing sentence stems, generating lesson ideas, and analyzing data.

Here, the teachers can take the initial AI-generated content and revise it for factual inaccuracies. They might also add elements based on their local context and knowledge of students. In other words, they take the “vanilla” of AI and make it their own:

With this approach, teachers can use AI to plan and prep but not as a teaching method. Here the AI tools function primarily as a time-saving or even money-saving mechanism. Meanwhile, students have no access to AI whatsoever. In some cases, however, teachers might incorporate AI as a time-saving device only when it does not connect with the specific learning outcomes at hand. A key element here is that the learning drives the AI integration and not the other way around.

What does this look like?

In a PBL unit, this would most likely look like an AI-Resistant unit with a few modifications. For example, a teacher might use AI tools to brainstorm some ideas for the project itself. After creating an initial outline, the teacher might ask an AI tool to design an initial project one-pager explaining the entire process, which the teacher then edits. This teacher might use an AI platform to design sentence stems, front-loaded vocabulary, and even slideshows for a few of the direct instruction embedded throughout the project. Perhaps even add a few AI-generated tutorials or handouts for students who need extra academic support throughout the project. Again, the AI would function primarily as a way to make differentiated instruction more feasible throughout the project.

For students writing, a teacher might begin with AI-resistant writing prompts described in the last section. But this same teacher might create some writing tutorials, such as an explanation of verb tense structures for ELL students or an explanation of how to use supporting details to back up a claim. Again, this teacher might create skill practice handouts, sentence stems, or leveled readers that students can access. However, on the student end, it would look fairly AI-resistant.

A class discussion would be in-person and dynamic but the teacher might use AI to help support the discussion process. The teacher might craft a quick inquiry-based activity where students are given an initial question (what makes fireworks change colors), followed by a guided discussion where they come up with a theory, then a short informational text explanation, and a whole class discussion on where they were right or wrong. Here, the teacher uses AI platforms to design the leveled readers and the discussion stems for students.

In a choice-based activity, this teacher might go with an advanced choice menu, where students click on a learning target they need to master and it takes them to a curated set of resources (some online, some generated by AI) followed by a set of 3-5 assignment choices. Again, a team of teachers might create an initial set of resources that can be modified based on what they know about their students.

 

3. The AI-Integrated Approach

This third approach focuses on the question, “What does it mean to use AI ethically and wisely?” and then attempts to integrate AI into the curriculum in a way that is human-driven but AI-informed. This is a blended approach that is driven by the learning targets rather than the technology. However, students and teachers use AI platforms as a way to improve the learning outcomes without falling into cognitive atrophy (where the AI does too much of the mental load).

With this blended approach both students and teachers use AI tools to save time on the repetitive tasks that aren’t central to the standards they are learning. A student might use an AI tool to do jump cuts in video or to create images for a slideshow presentation. But they might also use AI tools to build background knowledge of conceptual understanding. Here, a student can ask the AI chatbot a series of personalized questions or ask for feedback on a piece of writing. In this sense, the AI tools function as thought partners for feedback, co-creation, curiosity, and personalized scaffolding. I’ve even seen how students with executive function challenges have used chatbots to estimate time, break down tasks, and reflect on their progress.

What does this look like?

In a PBL unit, teachers might use AI for the back-end elements described in the previous section. However, students can also use AI tools throughout the project-based learning process. In the inquiry phase, students might do a question and answer session with AI chatbots to build their background knowledge. During online research, they might use an AI tool to change the reading level of a text or clarify a misunderstanding. They might ask for a summary before doing a deep dive into reading an article. If they’re doing quantitative research, students might use AI to help analyze and visualize data. During ideation, students can use AI tools as an additional group member who provides ideas or they might ask it to do a thought experiment like a “pre-mortem” where the generative AI finds challenges or flaws ahead of time within the design. They can also use AI to get feedback on their work or to help them stay on task through the project management process.

An AI-integrated approach to writing might include a student question and answer session with a chatbot to build conceptual understanding. A student might use the same chatbot during research to help clarify misunderstandings or as a reflection partner, where the chatbot asks clarifying and summarizing questions to the student to help improve retention and recall. This same student would then use a web or sketchnote followed by an outline. This student could then ask the AI to generate an outline and compare the human designed and AI generated outlines to see what things they would want to modify. For the most part, the student would generate the writing from scratch but might also use AI to create a paragraph that can then be heavily revised using a color-coded system.

In a class discussion, students using an integrated approach might use an AI tool for providing sentence stems in real time. Or they might use a fishbowl technique for a Socratic Seminar but then use an AI to ask their own questions or engage in fact-checking with a tool like Consensus. Another option might be a small group discussion where the team discusses a concept while they are all sharing a single computer with a chatbot. A group of fifth grade students could work together to interview a chatbot who takes on the role of a fictional woman from Sparta and woman from Athens. Or they might allow the chatbot to ask them questions as they test their own theories for why the sky changes color or how the pyramids were made.

For choice-based learning, an AI-Integrated approach would focus on personalized learning rather than adaptive learning:

 

4. The AI-Driven Approach

An AI-Driven Approach uses AI to transform the learning process entirely. If you imagine the SAMR Model, this would be that top level of Redefinition. While many of the previous approaches focus on using the learning targets to drive the AI usage, this approach recognizes that the technology itself is constantly changing both what we learn and the way we learn. There’s this ongoing relationship between technology and the learning standards.

If this sounds too tech-centric, consider how controversial it was when Ivy League universities began to de-emphasize oral exams and memorization for essays and multiple choice tests. The technology shaped the way they learned. Eventually, teachers placed more emphasis on the craft of writing, including the role of research. Or, more recently, when the US adopted the Common Core Standards, we added specific standards for online research and analysis of multimedia. In addition, technology continues to reshape how experts in various fields approach their subject matter (how historians, mathematicians, or scientists approach their methodology). Meanwhile, as machine learning continues to redefine every industry, we will continue ask, “What exactly are we preparing students for?”

Machine learning will continue to transform what students learn and how they learn it. This AI-Driven approach moves away from the question, “How do we keep AI away from this course content?” or even “How can we integrate AI into this existing content?” and instead asks, “How can we re-imagine this course for a world of machine learning? What are some of the powerful capabilities of these tools and how do we leverage it in an ethical way?”

What does this look like?

An AI-Driven approach might involve a STEM course where students use AI to simulate massive amounts of scenarios so that students can solve “sticky problems” where the solution leads to new challenges. Or they might use a combination of virtual reality and AI to create three models in engineering or to do a simulated electron microscope. Again, using AI with virtual reality, students could potentially visit key historical landmarks.

In a PBL unit, an AI-Driven approach might actually include a fourth group member who is an AI chatbot functioning as a coach or project manager. Students would use AI to analyze massive data sets and look trends during research. They might use AI as an ongoing co-creation tool, where they take the AI-generated content (computer code, video, image, text) and then heavily modify it to make it their own.

What about writing? Here’s where I like to think about music for a moment. Drum kits and synthesizers certainly allowed artists to replace a live musician with something more predictable and precise. But ultimately, these tools also allowed artists to re-imagine the musical experience and forge new pathways with entirely new genres. I might not be a fan of most EDM music but I can’t deny the human creativity involved in using these musical tools in innovative ways. Similarly, there is some real possibility in students using AI tools as a new form of composition and potentially redefining what it means to write. It will likely ruffle some feathers and I am firmly in the camp of students learning to write from scratch (particularly because of the notion of learning through writing). But I can also see how AI-generated text might lead to students thinking more divergently as they take the bland text and make it both more creative and more contextual.

In terms of choice-based learning, an AI-Driven approach would be fully adaptive, with the AI generating new content, answering questions, and providing additional questions at a fully individualized level. With this approach, the AI could adapt to student curiosity and interests but also to student energy level, motivation, and efficacy. One student might need shorter, more frequent feedback while another needs less frequent feedback with deeper explanations.

 

Who Gets to Decide?

Ultimately, all stakeholders should have a voice in where and how we use AI within our institutions. This includes the students, the teachers, the leadership, the coaches, but also the larger community. We need to have conversations with subject area and industry experts. We need to invite computer science and human psychology experts alike. In sharing this framework, my hope is that we can have a visual continuum where we can talk about what we think is most ethical and appropriate in any given circumstance.

 

Shifting Between Approaches

Note that these four models can work as school-wide approaches and philosophies. A Montessori school, for example, will likely choose to be fully AI-Resistant. Meanwhile, a tech-centered entrepreneurial school might take an AI-Driven approach and ask, “What are the limits of AI? How can we use it to change the entire learning process?”

We can also see these approaches play out in a course by course or grade by grade level. You might have a fully AI-Resistant Theater or PE course but go AI-Integrated in a Calculus course. You might avoid using AI with younger students but gradually integrate AI into the learning process in upper elementary or middle school.

You might switch between these models on an assignment by assignment bases or even within a single lesson. In my assessment course, I had students complete an in-class AI-Resistant task, then switch to something AI-Integrated, then move toward a homework assignment that was much more AI-Driven.

 

What About Consistency?

I recently shared this sliding scale in a workshop with university faculty members from multiple institutions. I had professors use sticky notes to consider which tasks would fit into each approach in the courses that they taught. It was fascinating to see the diversity of approaches across so many disciplines.

Afterward, I walked them through a See-Think-Wonder.

“This looks messy and inconsistent,” someone shared.

“I agree, it sends the message to students that we don’t have a unified approach to AI, much less a standardized policy,” another professor added. “I’m leading the committee trying to forge a single, agreeable policy, and I can’t imagine how we deal with this.”

“What if flexibility is the policy?” someone chimed in. “What if we gave people the permission to move between these approaches in an ongoing sliding scale?”

Another professor pointed to the sticky notes. “I would hate to be a student trying to navigate such conflicting messages. Imagine being in one class where you have AI-driven simulations and they’re making apps with AI and they’re getting feedback with AI and then another class has blue books.”

“Okay, but just hear me out. Isn’t that complexity exactly what students will need once they leave our institutions? They’re going to be navigating complex systems and changing expectations. They’re going to be needing to work on a task that is fully human and AI-resistant and then shift into an integration mode for an entire project,” another professor pointed out.

“But how do we provide this type of complexity and intentionality but also some consistent expectations?” someone asked.

I don’t think there’s an easy answer to this but I also wonder if this is part of a larger tension between educator autonomy and educational expectations and between the need for consistency and the need for complexity. As educators, we inhabit a world of paradox, where ideals and expectations are constantly in tension.

True, it creates confusion and even complexity. But it also creates nuance and beauty and, if we’re open to it, humility. I realize the question, “how should we approach AI?” might seem timely. But I also think it’s timeless.

Get the FREE eBook!

Subscribe to my newsletter and get the  A Beginner’s Guide to Artificial Intelligence in the Education. You can also check out other articles, videos, and podcasts in my AI for Education Hub.

Fill out the form below to access the FREE eBook: