The Challenge
Our Approach
We run an iterative research and development cycle, with educators, researchers, and engineers as equal partners at every step. We don't start by asking what AI can do. We start by asking what teachers and students need. Then educators, researchers, and engineers work together to identify what's missing, build it, and test it in real classrooms. What we learn informs the next turn of the cycle.
01
AI can play many distinct roles in a classroom, and each requires its own standards, evidence, and safeguards. With each group of teachers and students, we identify a specific role AI could play to meet a real need, and what teachers uniquely bring alongside it. Then we study what that role requires, where its limits are, and what bar it must meet.
03
Teachers, researchers, engineers, and designers build classroom tools on those capabilities, along with new ways of teaching alongside them. They then test our work where it counts: in real classrooms.
02
Much of the technology needed for AI to play these roles well doesn’t exist off the shelf. We build the missing layer between frontier AI models and classroom apps — validated capabilities that tools can build on, like computational methods that help AI measure student skills or track how concepts connect across a semester.
04
Classroom data tells us whether the AI roles we selected and the capabilities and tools we built actually helped. Those lessons sharpen our hypotheses, strengthen our capabilities and tools, and inform our next cycle of R&D.
01
AI can play many distinct roles in a classroom, and each requires its own standards, evidence, and safeguards. With each group of teachers and students, we identify a specific role AI could play to meet a real need — and what teachers uniquely bring alongside it.
02
Much of the technology needed for AI to play these roles well doesn’t exist off the shelf. We build the missing layer between frontier AI models and classroom apps: validated capabilities that tools can build on.
03
Teachers, researchers, engineers, and designers build classroom tools on those capabilities, along with new ways of teaching alongside them. Then they test the work where it counts: in real classrooms.
04
Classroom data tells us whether the role we chose and the capabilities we built actually helped. Those lessons sharpen our hypotheses and inform the next turn of the cycle.
Each co-design cycle produces more than prototypes. It creates evidence about what students need, what AI can do reliably, and what teachers need to use it well. AugmentED turns those lessons into research and technical foundations the field can trust and build on.
Ideas and evidence that help the field reason about AI in education: research insights on the roles AI should (and shouldn’t) play, methods for evaluating AI, and tested models for teaching and learning alongside it.
Reusable infrastructure that future classroom tools can be built on, from validated AI capabilities to the datasets that power them. We build these to generalize and recombine to serve many subjects and types of schools.
Photography to supply — a student mapping connected ideas at a whiteboard
Tested classroom resources that help educators apply AI, including AI-powered applications, implementation guides, and proven methods for teaching alongside AI, all built with teachers to solve real challenges in their classrooms.
Our Current Work
In our first cohort, teachers from several pioneering high schools worked with researchers and engineers to co-design AI-powered tools. Each project below is a work in progress—and each one feeds reusable capabilities back into a foundation that others can build on.
All three put the same AI role—AI as cognitive extender—to work. For teachers and students, the AI surfaces patterns hidden in more information than any person could track alone.
Photography to supply — a student recording a spoken reflection
Museum High School
During multi-week projects, students respond to teacher prompts with short voice memos. The tool surfaces where each student and group is progressing, stuck, or drifting off track, so teachers can provide specific, meaningful feedback for every student.
Photography to supply — four students working on a group project
High Tech High Mesa & High Tech High International
After students reflect on their small-group work, the tool generates insights about how each group is functioning. It makes group dynamics visible to teachers in real time, so they can coach students on collaboration and repair problems before they harden.
Photography to supply — a student mapping connected ideas at a whiteboard
Cross Town High
Students learn more deeply when new material connects to what they already know and care about. This tool maps a teacher’s course material to each student’s knowledge, experiences, and interests and then suggests personalized connections.
Our Thinking
With our research partners, we study what each role AI can play in a classroom makes possible, where it falls short, and what we can build to push those limits. Three of our recent papers:
Can AI assess a skill as complex as critical thinking? We find that carefully scaffolded language models can measure some critical-thinking subskills in student writing, and we map where they still fall short.
We share a standardized, step-by-step process for improving and validating the capacity of AI systems to detect complex skills like critical thinking, so the field can build measurement tools without starting from scratch each time.
We define five distinct roles AI can play in the classroom, and show how a tool's role shapes what it should do, how to judge whether it works, and what safeguards it needs.
We co-design with educators, publish what we learn, and build infrastructure for the whole field. If that's the future of AI in education you want, come build it with us.
AI is arriving in classrooms whether schools are ready or not. The danger is that some are rushing in without asking what AI can do well, what teachers uniquely bring, or what students actually need.
Photography to supply — a teacher and students working together in a classroom
Most of us built our critical thinking, writing ability, and judgment without AI, which is what allows us to use it well now. We direct it rather than defer to it. Students who lean on AI too early can short-circuit those skills before they form.
It's a safe bet that AI can improve how people learn. But nobody has discovered how. The entire field fundamentally lacks research into the right ways to integrate AI into the classroom.
Ed tech companies build the apps teachers and students use. The big AI labs build the underlying models that power these apps. What's missing is the layer in between: the foundational capabilities that let AI understand what a student knows, measure skills as complex as critical thinking, and adapt to what's happening in a real classroom.
Photography to supply — a researcher and engineer reviewing a prototype
Photography to supply — a teacher working alongside two students
Parents and teachers are alarmed. Some districts are pushing screens and AI out of schools entirely. This instinct is understandable, but both extremes carry real costs. Letting AI take over the classroom would stunt student learning and hollow out the relationships it depends on. Yet pushing AI out altogether leaves students unprepared for a world that already runs on it.
The gap exists because frontier AI and classroom learning operate on different timelines and constraints. Our work closes that gap by creating the infrastructure that lets them work together—safely, effectively, and in service of what students actually need to learn.
Photography to supply — a wide shot of a classroom in session
We believe better educational AI will emerge from discovering what classrooms actually need, building solutions with real educators and students, and testing them in real classrooms. What we learn from that testing flows back into better capabilities and tools. That's why every AugmentED project combines research and development in a iterative co-design cycle.
Education technology companies build the applications teachers and students use every day. The large AI labs build the underlying models. What's missing is the layer in between—the capabilities that let AI understand what a student knows, measure skills as complex as critical thinking, and adapt to what's happening in a real classroom. That's the layer we are building.
01 Define the role
We start by asking, not what AI can do, but what teachers and students need. At the start of each co-design cycle, our teacher and research partners define a role AI can play to meet a real classroom need, such as enhancing a teacher's understanding of her students' prior experiences and interests, assessing complex skills, or facilitating feedback. That role becomes the North Star for everything that follows.
Photography to supply — an engineer and a teacher reviewing a tool together
02 Build the capabilities
Our educators, researchers, and engineers then build the underlying infrastructure: the reusable technical capabilities a tool needs to play its chosen role well. For the roles we're currently exploring, that includes building capabilities such as validated ways for AI to measure and support durable skills and a living map of how the ideas in a particular teacher's class connect to each other and to students' prior experiences and interests. Most of these capabilities don't exist yet. Once we build and prove them, they can be reused, adapted, and made available to others to build on.
03 Co-design the applications
Our interdisciplinary teams—teachers, researchers, engineers, and designers working as true partners—use that infrastructure to build and test AI-powered tools, and new ways of teaching alongside them. This second part is crucial: an AI tool might help students evaluate sources or collaboratively solve problems, but it's only truly effective when paired with a teaching approach that combines what AI and human teachers each do best.
Photography to supply — four students working on a group project
04 Test, learn, begin again.
Every co-design cycle generates evidence about what AI should do in education, when it works, and why. We openly share our research findings, evaluation methods, and models for teaching and learning alongside AI, so the field can build from evidence instead of assumptions.
Behind every successful classroom tool are foundational capabilities that make it possible. We develop and share reusable capabilities, from specially trained models and datasets to ways of representing what's being taught and learned in a classroom, that any educational AI application can build on.
Research and capabilities only matter if they improve learning. We build and validate classroom tools and teaching methods alongside educators and students to demonstrate what works in practice.
01
Define the role
02
Build the capabilities
03
Co-design the applications
04
Test, learn, begin again.
Our Approach
What we learn informs the next turn of the cycle.
Step 01
AI can play many distinct roles in a classroom, and each requires its own standards, evidence, and safeguards.
Step 02
We build the missing layer between frontier AI models and classroom apps — validated capabilities that tools can build on.
Step 03
Teachers, researchers, engineers, and designers build classroom tools on those capabilities, then test them where it counts.
Step 04
Classroom data tells us whether what we built actually helped. Those lessons inform our next cycle of R&D.
01
AI can play many distinct roles in a classroom, and each requires its own standards, evidence, and safeguards.
02
We build the missing layer between frontier AI models and classroom apps — validated capabilities that tools can build on.
03
Teachers, researchers, engineers, and designers build classroom tools on those capabilities, then test them where it counts.
04
Classroom data tells us whether what we built actually helped. Those lessons inform our next cycle of R&D.
Photography to supply — educators and a researcher in a co-design workshop
The Feedback Loop
Every application is tested in real classrooms alongside educators and students. What we learn tells us which capabilities to build or improve next, while new and improved capabilities make better applications possible. Together, they form a continuous research and development cycle.
Each iteration of the cycle strengthens the field's understanding of what roles AI should play, what capabilities those roles require, and how those ideas translate into practical tools that genuinely improve learning.
AugmentED brings together people from classrooms, research labs, and engineering teams who share a conviction that AI should augment human teaching, not replace it. We build the infrastructure that lets teachers and students work alongside frontier AI in ways that deepen their thinking, learning, and relationships rather than short-circuit them.
People who believe AI amplifies rather than replaces human thinking.
We collaborated with XQ and High Tech High to identify exceptional educators for our first co-design teams. Our inaugural AugmentED Education Fellowship cohort includes teachers from Crosstown High School (Memphis, TN), Grand Rapids Public Museum High School (Grand Rapids, MI), Thomas A. Edison Career and Technical Education High School (Queens, NY), and several High Tech High campuses in the San Diego region. We've selected these schools for their commitment to transforming the high school experience for students—the majority of whom are students of color and economically disadvantaged—by making learning meaningful and engaging, and preparing them with the critical skills they need to thrive in the age of AI.
We co-design with educators, publish what we learn, and build infrastructure for the whole field.
Connect
Get updates on our work and research findings.
