The Problem With Designing for the “Average” Learner
At a recent conference, Dr. Sarah Quinn—who teaches future educators how to work with neurodiverse learners—shared a story that’s stuck with me. In the 1950s, the U.S. Air Force measured thousands of pilots across 140 dimensions—everything from arm length to shoulder width. They averaged those numbers to design the “perfect” cockpit for the “average” pilot. The result? Not a single pilot actually fit the design. The cockpit built for the aggregate failed to fit anyone at all.
Real safety depended on designing for variability, not averages. Education faces the same trap. We continue to build curricula for the “average student,” even though no such student exists. Even frameworks like Universal Design for Learning (UDL) emphasize variability, but they still depend on knowing enough about each student to apply the right adaptations.
Without individualized insight, curriculum defaults to the aggregate—and learners get left behind.
From Free Energy to Freeing Teachers
When I collaborated with neuroscientist Karl Friston—the mind behind the Free Energy Principle—on real-world applications of his theory, what struck me wasn’t just the math. It was the biological truth: organisms learn by taking in sensory signals, building a model, and then acting on it. Adaptive learning should work the same way—gathering signals from students, building models of their strengths and needs, and then acting with purpose.
Traditional textbooks and static curricula skip that modeling step. They act on assumptions about “typical learners.” AI can help fill the gap. By continuously modeling learners, AI can take on the invisible but heavy lift of differentiation, giving teachers more time and energy for human connection.
Smarter Assessments: Modeling the Learner in Real Time
I collaborated with Dr. Jason Yeatman, who is on a mission to help educators “understand and support the diversity of learners.” His work on the Rapid Online Assessment of Reading (ROAR) reduces testing time to just a few minutes while still showing whether a child has reached the critical reading decoding threshold. We integrated ROAR into a literacy application so lessons could directly target decoding gaps. It was inspiring to see neuroscience research turn into a tool that helped teachers know exactly where to focus.
At Everway, we’re taking this further with adaptive assessments that mirror biological feedback loops. Like computer-adaptive tests, each question dynamically adjusts based on the learner’s previous response. This produces a more efficient, evolving model of each student’s strengths and needs.
Beyond Tests: Mining Existing Data
Schools often over-test because assessments double as accountability tools for districts and states. But adaptive curriculum doesn’t always require new tests. Instead, AI can model learners by analyzing signals that already exist: assessment histories, activity logs, IEP documentation, and even gameplay data.
At Odeum, for example, we capture every learner interaction within immersive quests. If a student in a Revolutionary War simulation shows confusion about the Stamp Act, the system adapts—triggering side quests or scaffolding aligned to their needs. Each action becomes a “signal” that refines the learner model in real time. The goal is to let learners experience content, while the system quietly observes and adapts.
The Responsibility of Modeling Students
Collecting learner data brings responsibility. There’s a difference between pre-interpreted data (like “this student is on the spectrum”) and raw signals (like “this student struggles with short-term memory tasks”). The former risks generic, reductive conclusions; the latter enables nuanced, strength-based personalization.
Adaptive systems must avoid labeling learners broadly and instead focus on specific traits that impact learning. Knowing a student processes information visually faster than verbally, for example, can guide more equitable content delivery without stereotyping or bias. This is the kind of careful modeling work that must sit at the heart of adaptive education.
Turning Models Into Action
Once we’ve modeled the learner, the next challenge is adaptation. If a student learning math hasn’t met the reading decoding threshold but thrives with visual supports and loves space exploration, AI should recommend content scaffolded with visuals and wrapped in a space-themed narrative.
This requires human expertise. At AERDF’s Reading Reimagined, I worked alongside researchers actively shaping best practices for literacy instruction. At Odeum, we collaborate with language acquisition experts. At Everway, our instructional designers and teachers ensure that AI-generated adaptations remain aligned with standards, pedagogy, and IEP goals.
Domain experts provide the compass that guides AI’s power. Without them, adaptive systems risk being clever at personalization but shallow on pedagogy.
Measuring Mastery in Personalized Ways
Instructional delivery is only half of adaptive learning. The other half is how learners demonstrate mastery. If we know a student has low visual processing but strong auditory skills, testing their understanding through spoken response will produce more accurate results than a visual-heavy exam.
AI allows us to match assessment methods to learner strengths, ensuring that we measure true understanding rather than penalizing students for unrelated challenges. Done well, this creates a more equitable path to demonstrating mastery.
Freeing Teachers to Do What They Do Best
At Everway, we are working toward reducing the cognitive load teachers face in planning and differentiation so they can focus on their students. The aim is to give educators a clear, accurate picture of where each student is—without hours of grading and preparation. AI becomes the co-teacher that works behind the scenes.
At Odeum, I’ve seen similar results in immersive learning: when a system adapts a quest or challenge automatically, teachers can step back and focus on the spark in the student’s eyes. In both cases, AI doesn’t replace teachers—it frees them to be who they wanted to be when they entered the profession in the first place: mentors, guides, and inspirers of curiosity.
Moving Beyond the Mythical “Average”
Just as the Air Force learned there was no “average pilot,” we must accept there is no “average student.” By embracing AI-powered adaptive learning, we can finally move beyond teaching to the aggregate and begin teaching to the individual.
With careful attention to data ethics, pedagogy, and teacher empowerment, we can create curriculum that truly learns back from every learner. That is how education becomes both more human and more effective—by designing for the diversity that has always been there.
