In the early 1950s, the US Air Force had a serious problem. Their pilots were crashing planes at an alarming rate, yet no mechanical defects could be found. The issue, it turned out, wasn’t the pilots’ skills or the planes’ engines. It was the cockpit.
The cockpits were designed for the "average" pilot, based on the physical dimensions of thousands of airmen. The assumption was that if you designed for the average, you’d accommodate the majority. But when researchers measured over 4,000 pilots against ten specific physical dimensions, they found something startling: not a single pilot fit the average on all ten counts banning the average in education. Zero.
By designing for the average, the Air Force had designed for no one. Their solution was radical but simple: ban the average. They introduced adjustable seats, pedals, and helmet straps. Pilot performance soared.
For decades, our education system has been stuck in that pre-1950s cockpit. We’ve built a massive infrastructure based on the mathematical abstraction of the "average student." But thanks to the rise of artificial intelligence, we are finally seeing the end of average education. We are entering an era where the "adjustable seat" is digital, scalable, and capable of vindicating the unique shape of every human mind.
The Myth of Average and the Jagged Learning Profile
The concept of the average human is a statistical convenience, not a biological reality. This is the core of the "Jaggedness Principle," popularized by education researcher Todd Rose. Human talent isn’t a single, sliding scale where you are either "smart" (high IQ) or "not smart" (low IQ). Instead, our abilities are multi-dimensional and weakly correlated multi-dimensional human talent.
This creates what is known as a jagged learning profile. A learner might possess exceptional conceptual reasoning (a peak) but struggle significantly with working memory or processing speed (a trough). In a standard classroom, these nuances are flattened.
Consider a student with high verbal aptitude but slow processing speed. In a timed essay exam—a staple of the factory model of education—this student fails. The system records this as a lack of understanding, when in reality, it’s a measurement error. The assessment measured their speed, not their knowledge. Standard metrics like GPAs aggregate these peaks and valleys into a single number, effectively erasing individual student nuances.
For neurodivergent learners, these profiles are often described as "spiky," with profound disparities between strengths and challenges strengths of spiky profiles. But the truth is, we are all spiky to some degree. The traditional system just forces us to hide our spikes or fall through the cracks created by our valleys.
Why the Factory Model is Breaking
Our current educational infrastructure was built for an industrial age. It relies on age-based cohorts and standardized timelines. The underlying assumption is that faster processing equals higher intelligence and that all students should master content at standardized educational timelines.
This creates a "one-size-fits-none" environment. When a curriculum is rigid:
- The Gifted Disengage: Students ready to race ahead are capped by the class average, leading to boredom and addressing student talent loss.
- The Struggling Fall Behind: Students who need just a little more time on a specific concept are pushed forward regardless, accumulating gaps in their foundational knowledge.
In this system, a jagged learning profile is a liability. High talent in one area becomes irrelevant if you can't meet the average requirements in another. But this is where adaptive AI learning systems are changing the game.
AI as the Vindicator: The Adjustable Seat for the Mind
Artificial Intelligence offers the first scalable mechanism to operationalize the Jaggedness Principle. AI-driven platforms act as the adjustable seat, decoupling learning progress from rigid time constraints and standardized delivery methods.
1. Decoupling Progress from Time
In a traditional classroom, time is constant, and learning is variable. You get two weeks for algebra; if you get a C, the class moves on. AI flips this: learning becomes constant, and time becomes variable.
Platforms like ALEKS (Assessment and LEarning in Knowledge Spaces) use complex algorithms to map a student's precise knowledge state. Rather than forcing a student through a linear textbook, the system identifies exactly what a student knows, what they don't know, and—crucially—what they are ready to learn next ALEKS knowledge mapping.
This approach has shown quantifiable success. At Harper College, implementing ALEKS for math remediation led to a 19% increase in students passing college-level courses and significantly boosted student retention and pass rates. By surgically repairing specific knowledge gaps rather than forcing students to retake entire courses, the system honors the learner's jagged profile.
2. Scaffolding Weaknesses
Generative AI tools are evolving into cognitive prosthetics that support executive function and processing challenges. This allows learners to demonstrate their true capacity without being bottlenecked by their weaker skills.
Take Goblin Tools, for example. It utilizes AI to break down overwhelming tasks into manageable micro-steps or adjust the tone of communication. For a learner with ADHD who understands how to do a task but struggles to initiate it, this tool bridges the gap between capability and task execution. Similarly, tools like Scholarcy can summarize dense academic texts into flashcards, allowing students with dyslexia or slower processing speeds to access high-level concepts without being overwhelmed by dense reading volume.
3. Socratic Tutoring
The fear that AI will make students passive is being countered by the next generation of AI tutors, such as Khan Academy’s Khanmigo. Instead of simply providing answers, these tools use a Socratic approach. They ask probing questions like, "What do you think the next step is?" to guide the thinking process with Socratic AI tutors.
This provides a safe, non-judgmental space for remediation. A student can ask "stupid" questions or ask for a concept to be explained "like I play Minecraft" without the social stigma of raising their hand in a crowded room. It allows the environment to fit the student, not the other way around.
Navigating the "Jagged Frontier" of AI
As we integrate these tools, it’s vital for learners to understand that AI itself has a jagged profile. Wharton professor Ethan Mollick describes this as the "Jagged Frontier" of AI capabilities. Just like humans, AI models have peaks and troughs; they might excel at complex creative writing but fail at simple arithmetic failures in AI logic.
What does this mean for learners? Success in the future won't just be about knowing facts; it will be about knowing how to partner with AI to smooth out your own jagged edges. Mollick identifies two successful models for this interaction:
- Centaurs: Humans who clearly divide tasks, handling the strategy while the AI handles the execution (e.g., coding).
- Cyborgs: Humans who integrate AI deeply into their workflow, moving back and forth fluid human-AI collaboration.
By teaching students to be Centaurs or Cyborgs, we empower them to leverage AI to mitigate their valleys while amplifying their peaks. This effectively "levels up" their capabilities, allowing a student with great ideas but poor grammar, or great logic but poor syntax, to compete on the merit of their thinking.
Challenges and Critical Perspectives
While the "end of average" is promising, we must approach this shift with nuance. There is a genuine risk of the "Hollowed Mind" or the "Sovereignty Trap." If AI automates too much of the cognitive struggle, students may bypass the "desirable difficulties" required for deep learning preserving cognitive sovereignty. We must ensure that the "adjustable seat" supports the pilot, rather than flying the plane for them.
Furthermore, we must confront the digital divide. The adjustable seat works, but only for those who can afford the cockpit. There is a risk that wealthy students will become "Cyborgs" AI-enhanced learners with AI superpowers, while under-resourced students remain stuck in the factory model. To truly kill the average, these adaptive tools must be accessible public goods, not luxury add-ons.
Conclusion: A New Architecture for Potential
The death of the average in education is long overdue. For too long, we have blamed the pilots for crashing when the cockpit was never designed to fit them.
We are moving toward a future where "falling behind" is an obsolete concept, replaced by asynchronous mastery. In this new landscape, a jagged learning profile is not a defect to be fixed—it is a unique architecture to be supported. By leveraging adaptive AI learning systems, we can finally build an educational environment that treats personalization not as a luxury, but as a geometric necessity for human cognition.
For learners, the message is clear: Your "spiky" profile is no longer a liability. With the right tools, it is your greatest asset.