The two frameworks most districts still run on weren’t built for an AI classroom. Here’s a new synthesis — five domains that describe what actually happens when humans and AI learn together.
Every district I talk to is being sold the same promise: buy this tool, and AI will make teaching easier. Six months later, most teachers have quietly stopped opening it — and the ones who kept going are asking a sharper question than any product demo answers. Not which tool? but what are we actually teaching students to do, now that a machine can produce a five-paragraph essay in nine seconds?
That question doesn’t have a software answer. It has a learning answer — and the two frameworks most districts still run on weren’t built to give it.
Bloom’s Taxonomy — first published in 1956, revised in 2001 — put Remember at the base of the pyramid. The logic was sound for its time: you had to hold facts in your head before you could do anything interesting with them. Recall was the floor of learning, and everything higher-order was built on top of it.
Charlotte Danielson’s Framework for Teaching remains one of the most useful teacher-evaluation instruments ever written. Its emphasis on planning, classroom culture, assessment, and reflective practice is still exactly right. But its descriptors were written before a student could generate a full essay draft, a lab analysis, or a working block of code from a single sentence. Neither framework was designed to evaluate a human–AI learning partnership — because when they were written, there wasn’t one.
I don’t think the answer is to throw either one out. I think the floor has moved, and our frameworks haven’t caught up to where it is now.
I want to separate two things clearly, because the AI-in-education conversation constantly blurs them. There is what the evidence supports, and there is my own proposed framework built on top of it. This section is the evidence.
The World Economic Forum’s Future of Jobs Report 2025 found that analytical thinking is now the number-one core skill employers say they need, cited by seven in ten companies. The fastest-growing skills on their list are AI and big data, followed closely by creative thinking, curiosity and lifelong learning, and resilience. And employers expect 39% of workers’ core skills to change by 2030. The skills that are appreciating in value are the ones that sit at the top of Bloom’s pyramid, not the bottom.
The OECD’s Learning Compass 2030 centers what it calls transformative competencies — creating new value, reconciling tensions and dilemmas, and taking responsibility — and frames students as agents who navigate their learning rather than passengers who receive it. UNESCO’s 2023 Guidance for Generative AI in Education and Research argues for a human-centered approach, with real guardrails: data-privacy protection and age limits on independent student use of generative AI, so the technology supports human judgment instead of quietly replacing it. And the ISTE Standards for Students already ask learners to be Knowledge Constructors who “critically curate” sources — evaluation was baked into the standards before ChatGPT existed.
Put together, the pattern is consistent: none of these bodies say memorization is obsolete. They say the center of gravity has moved — from holding information to interrogating it.
It would be easy to read all of this as “AI means students don’t need to know things anymore.” That is the wrong lesson, and the cognitive science is unambiguous. Daniel Willingham and decades of research make the point plainly: you cannot think critically about — or fact-check — a topic you don’t actually understand. Background knowledge stored in long-term memory is what lets a student notice that an AI just invented a citation, flattened a historical cause, or “explained” a text it clearly misread.
So the shift is not that memory stops mattering. It’s that rote retrieval of instantly-retrievable facts is no longer the bottleneck, while the durable knowledge that powers judgment matters more than ever. Any AI-age framework that tells teachers to stop building knowledge has misread the science. Mine doesn’t.
Rather than replace Bloom’s, I reinterpret each level for a classroom where retrieval is free and judgment is scarce. The cognitive levels don’t change — what changes is the verb that defines mastery at each one, and the question a student has to be able to answer honestly.
Memorize and recall facts from memory
Where can I locate information I can actually trust?
Demonstrate comprehension of concepts
Can I teach this in my own words, without the tool?
Use information in new situations
Can I solve a real problem with it?
Break information into parts, find patterns
Can I compare sources, patterns, and assumptions?
Make judgments based on criteria
Can I tell what the AI got right — and what it got wrong?
Produce new or original work
Can I make something uniquely valuable with AI, not just because of it?
Notice what didn’t survive the translation: Memorize. Not because memory stopped mattering — see above — but because memorizing retrievable facts is no longer where the learning happens. The verbs that rise in its place are the ones a machine can’t do for a student: question, prompt, verify, challenge, connect, refine, and teach.
Here is where I go past reinterpreting the classics. Danielson evaluates teaching. Bloom’s describes thinking. Neither was built to describe the thing actually happening in classrooms right now: a human and an AI learning together. That’s the gap the Rooted Learning Framework is meant to fill. It’s my synthesis — offered for districts and educators to adapt, argue with, and make their own — organized into five domains.
The teacher designs experiences AI cannot complete on a student’s behalf.
Students learn to work with AI as a tool they supervise, not an authority they defer to.
The teacher shifts from delivering information to coaching judgment.
The capacities AI cannot replace become explicit, taught learning goals — not things we hope happen by accident.
Assessment moves from “What answer did you get?” to “How did you think?” — which is exactly the direction Danielson always pointed assessment, extended into an AI-supported classroom.
Danielson’s detailed descriptors are a strength for evaluators and an obstacle for busy teachers. So the framework uses a single four-level continuum, applied the same way across all five domains — practical enough to actually use in a walkthrough or a planning conversation.
AI is used mainly for efficiency — saving the adult time.
The starting point. AI is present but not yet integrated into learning design.
Students use AI occasionally, with heavy guidance.
AI use is intentional but still teacher-directed. Students are learning the guardrails.
Students use AI critically, ethically, and on purpose — consistently.
The target for most classrooms. Students demonstrate judgment alongside tool use.
Students independently use AI to solve authentic problems and create original work.
Students reflect on their thinking, improve their communities, and teach others.
This is the thinking underneath the curriculum-and-learning-systems audits I run and the classroom tools I build. When I sit with a district, the assessment question is never “which platform?” — it’s whether the district can see how students are thinking, not just which answers they landed on. That’s the same shift Domain 5 describes, and it’s the design principle behind AssessAlign, the standards-based formative assessment tool I’m building: capture the reasoning, not only the result.
A framework won’t configure your tech stack or write your ELD overlays. But it will keep you from spending a Title III budget on a tool that automates the wrong thing. Get clear on what you’re teaching students to do in an AI classroom, and the technology decision gets a lot simpler.
Want to see where your current curriculum, assessment, and tools line up against the five domains? Download the free Rooted Learning Framework self-assessment, or book a 30-minute discovery call and we’ll map it against what your district is actually navigating.
— Michelle is a K–8 educator, WIDA/ELD specialist, Science of Reading–certified practitioner, and full-stack developer. She is the founder of The Rooted Learner and works with DMV districts to build standards-aligned, board- and state-approved instructional tools.
