SAMR and AI: Don’t get stuck on substitution

Sohan Choudhury | sohan@flintk12.com

Dec 21, 2023

Juxtaposition of warm classroom with redefined teaching embracing AI and cold boring classroom without technology.

Illustrated by AI, written by humans.

A breathtaking new AI development, every single day. That’s our new reality. From powerful open-source models that can run on your phone to breakthroughs in generative AI for new modalities like music, there’s no shortage of incredible demos to see and experience.

But while these demonstrations are great at piquing our curiosity and showcasing what’s possible, it’s often unclear how new developments in AI will translate into an impact on education.

The SAMR model for technology integration.

This is where we can turn to the SAMR model, which shows us the tiers of technology integration as they relate to supporting learning. The SAMR model was created in 2010 to explain the four tiers of online learning, and goes as follows:

Substitution: Technology acts as a direct substitute, with no functional change. For online learning, this was equivalent to a teacher replacing a paper handout with a digital PDF.
Augmentation: Technology doesn’t change the content, but enhances the delivery of that content. For example, making a quiz more fun with Kahoot.
Modification: This is where technology starts to change teaching practices. For online learning, an example would be using an LMS to unlock student-to-student and student-to-teacher communication outside of class.
Redefinition: This is where new things happen that weren't possible before to change how teachers teach and how students learn. For example, imagine a class field trip in VR to the Louvre.

How well can we apply SAMR to AI?

Well, one key difference is that with AI, substitution happened quickly. AI tools today can already do many tasks that we currently do, which requires us to modify SAMR.

When it comes to AI, substitution is just realization

I love Dr. Nick Jackson’s piece, “SAMR and AI Chatbots”, because it reclassifies the “substitution” component of SAMR as a realization of the capabilities of AI, as opposed to a simple enhancement to education. This reflects the reality that for many — especially students — the “aha” moment with generative AI was in realizing how it could be used to substitute work such as writing papers.

Illustration of a student using AI to substitute their work by generating an essay.

Now, are there benefits to using AI to substitute work? Certainly. For teachers, for example, AI can substitute repetitive tasks that may otherwise contribute to burnout. It’s no surprise that hundreds of thousands of teachers have been using AI to write letters of recommendation, create classroom materials like worksheets or rubrics, or even to draft emails to parents.

Substitution, however, isn’t the end goal. When we simply use AI to replace the set of tasks that we currently do, we’re failing to scratch the surface of its potential. In his “Back to School with AI” blog series, Eric Hudson breaks down the research on the topic:

"In his 2022 paper “The Turing Trap,” Erik Brynjolfsson of Stanford University argues that the real potential of AI lies not in its ability to automate human tasks (thus, replacing the need for humans), but rather in its ability to augment human’s abilities, to help us be better at what we can do now and to enable us to extend our abilities in ways we haven’t conceived of yet."

Infographic showing new tasks that humans can do with the help of machines, tasks that humans can do, and human tasks that machines can automate.

Unfortunately, the vast majority of AI tools being built for educators today are focused on the dark blue square — on automating human tasks with AI.

SAMR and the AI EdTech landscape today

The SAMR framework, showing how most AI EdTech companies are currently focusing on substitution.

Since the launch of ChatGPT a year ago, many new EdTech companies have been created to improve the ease of use of AI for educators. Most of these new tools are focused on using AI to save teachers time on pre-existing tasks via substitution and augmentation:

MagicSchool is used by teachers to save time on everything from creating rubrics to writing questions for assessments.
Brisk Teaching is a Chrome extension that teachers can use to create lesson plans or worksheets with the help of AI.

The same goes for AI features released by incumbent EdTech companies as well. For instance, LMS providers like Blackboard have released AI assistants specifically to help teachers with tasks like designing curricula and creating rubrics for students.

The substitution of pre-existing tasks is the most obvious use case of AI, especially because it mimics the most common ways that people have been using ChatGPT so far.

Not only is substitution a popular use case of AI currently, it can be seen as a safe and measured starting point for AI integration. With other technologies, such as online learning, the shift from enhancement to transformation of learning took over a decade to happen and is still an ongoing process.

With AI, however, we might be able to shift to transformation much quicker than we think.

Why Flint jumps straight to transformation

At Flint, we’re solely focused on building AI tools to transform education. But that wasn’t always the case. At first, we actually started with substitution.

The first version of Flint was an AI copilot for teachers — a tool that any K-12 teacher could use to create classroom activities, worksheets, or lesson plans. Here’s what our original website looked like:

A screenshot of the original version of the flintk12.com website, showing an AI teacher copilot that can be used to create activities, worksheets, and quizzes.

This was the idea that Jinseo and I launched Flint with, and we built a world-class AI copilot that was used by over 500 teachers.

After a month, however, we shifted gears. While it was great to use AI to save teachers time, we found that often the most passionate and driven teachers prefer to manually create materials for students, and AI could be seen as detracting from their craft.

More importantly, we realized that the product we had built wasn’t directly improving the learning experience for students. AI today can at most mimic the work that humans produce when it comes to worksheets and classroom materials, meaning that the quality of learning for students would at best be the same as before.

We had the nagging feeling that there was potential in using AI for personalized learning, which would be truly transformative if it worked. So we set out to explore.

Very quickly, we saw firsthand that schools were ripe for transformation, and that AI today can support transformation via personalized learning. We found that many longstanding challenges in education, such as Bloom’s 2 sigma problem, could now be addressed. Generative AI finally makes it possible to personalize learning and assessment for students in subjects ranging from Spanish to Computer Science.

Additionally, we found that there’s an innate ambition in teachers to innovate practices to help students learn better. Many educators are specifically curious about using AI to transform teaching and learning, but lack the tools to do it because most tools today are built for substitution.

The SAMR framework, showing the shift of Flint from substitution into transformation.

So, we went for transformation. We built Flint into the best way for teachers to use AI to personalize learning for students, and the use cases we’ve seen have blown us away. Here’s Justin Cerenzia, Chair of Teaching and Learning at The Episcopal Academy, summarizing his experience with rolling Flint out at the school:

"The feedback from our community about Flint has been overwhelmingly positive. What has stood out the most, echoed by both teachers and students, is how Flint encourages a different mode of thinking. It's not about rote memorization or simply getting through an assignment; Flint has an uncanny ability to compel individuals to engage with material on a cognitive level that's often missing from more traditional assignment types. It's one thing to introduce a new tool that people like, but it's another to introduce one that prompts new ways of thinking and adds meaningful value to the educational process."

"This isn't just another tool designed to automate tasks; it's a resource that amplifies the craft of teaching while simultaneously enriching the learning experience. Flint is more than just a tool—it's a partner in the classroom that aids in facilitating valuable retrieval practice through immediate feedback. In doing so, Flint doesn't just make our lives easier; it makes us more effective educators and helps students learn more effectively."

Justin Cerenzia, Chair of Teaching and Learning at The Episcopal Academy

The philosophy of transformation

“We co-evolve with our tools. We invent new tools, and then our tools change us.” — Jeff Bezos

You don’t have to love Jeff Bezos, but he’s prescient about the role of technology in society. The above is what he’s recently said on AI.

Generative AI is an incredible new tool at our disposal, and the greatest potential of this tool will be in doing new things that were previously out of reach. For education, that might be teachers using AI to create new types of conversational assignments for students.

Ironically, many of the problems caused by AI in education can only be reliably solved through expanding our use of the technology for good. For example, if students are using AI to write papers and given that it’s not possible to detect AI writing, the path forward is to redefine assessment — to use the same adaptability and personalization offered by AI to create new ways of measuring student performance.

As AI can increasingly do more and more of the work that we do, we should welcome automation but see it as a starting point. AI automating work affords us the freedom to innovate and transform what we do. If we don’t, we’ll end up in a productivity race against the machine, and I wouldn’t bet on us to win.