All Tirzah Talks

AI & EDUCATION · OPINION

The Cognitive Overload Myth: Why Going Back to Basics Will Cost Your Child Their Future

Politicians are calling for workbooks. Educators are banning laptops. The headlines blame screens for a generation's declining performance. And the science of how the brain actually learns has never been clearer, or more ignored.

Tirzah Elese Bagnulo

Tirzah Elese Bagnulo

Founder & CEO, ExelAcademy · 9 min read

Learners working alongside AI

The Misunderstanding Shaping a Generation of Policy

The headlines are doing what headlines do. Gen Z is the first generation less cognitively capable than their parents. A reckoning for devices in schools is overdue. We spent $30 billion on laptops and got a literacy crisis.

Thirty-five US states now have laws limiting phone use in classrooms. England made phone-free schools the legal default this year. France, Italy, China, the Netherlands: the move toward restriction is a global wave.

And underneath all of it runs a single, seductive assumption: that less technology means less cognitive overload, and less cognitive overload means more learning.

It is an understandable conclusion. It is also, if you look at the science carefully, almost exactly backwards.

Cognitive load theory, the genuine science behind the "overload" concern, was developed by educational psychologist John Sweller in the late 1980s. Its core insight is real and important: working memory has limits, and instruction that overwhelms those limits produces confusion rather than learning.

But here is what the theory actually says, and what its popularisers consistently leave out. Cognitive load comes in three distinct types:

  • Extraneous load (the bad kind) comes from poor instructional design: irrelevant information, cluttered presentation, unclear explanations.
  • Intrinsic load comes from the inherent complexity of what is being learned.
  • Germane load (the kind that actually builds lasting knowledge) is the productive mental effort of making sense of new material, connecting it to what you already know, and reconstructing it in memory.

The policy conversation has collapsed all three into one. It has decided that any complexity, any richness of stimulus, any tool that does more than a printed page is "cognitive overload." And the prescribed solution (strip it back, return to workbooks, remove the screens) does not reduce harmful cognitive load. It removes the conditions for germane load. It removes the conditions for thinking.

Cognitive overload is not caused by too much information. It is caused by information presented in ways the brain cannot organise. The answer is better design, not less ambition.

What the Back-to-Basics Movement Gets Right, and What It Misses

The frustration driving the back-to-basics movement is legitimate. Billions were spent putting devices in classrooms without first answering the most important question: what are they for, and how will they be used? Students ended up with laptops as glorified distraction machines. Tablets replaced textbooks, but not the thinking that good textbooks can provoke. Novelty was mistaken for pedagogy.

The right response to that failure is to demand better design, better implementation and better purpose behind the technology. Instead, we are retreating. We are pulling the tools and keeping the pedagogy, which was the problem all along.

The claim that Gen Z is "less cognitively capable" comes from neuroscientist Jared Cooney Horvath's testimony to the US Senate, and it laid the blame at the door of technology. But read the argument closely and a more specific picture emerges. Horvath himself does not reject technology. He calls for aligning tools with how learning actually works. The problem was never simply that children had screens. It was that the screens were handed over without a framework for using them to think. Passive consumption disguised as learning. Browsing mistaken for research. Watching a video mistaken for understanding it.

The Workbook Was Never the Point

There is a deep nostalgia in calls for a return to traditional methods. But the workbook was not a uniquely effective tool. It was the only tool widely available, in an era before we understood cognitive science, before we had evidence on what builds durable knowledge, before we had the technology to personalise instruction at scale.

What made the best traditional classrooms work (Socratic questioning, high expectations, effortful retrieval, spaced practice) had almost nothing to do with the physical format of the materials. It had everything to do with the quality of the cognitive demand placed on students. You can do that with a workbook. You can do it without one. And you can now do it better than any teacher working alone with thirty children ever could, if the technology is designed for thinking rather than passive reception.

What the Evidence Actually Shows

The research on AI-supported learning is growing fast, and its direction is encouraging.

A 2025 review of 21 empirical studies comparing AI-driven tools with traditional methods found performance gains ranging from 15% to 35%, alongside faster task completion and higher learner satisfaction. In a 2025 randomised trial at Harvard, students taught physics by a carefully designed AI tutor learned more than twice as much as peers in an active-learning class, and did it in less time. A separate review of personalised adaptive learning found that 41 of 69 studies (59%) reported improved academic performance, while the rest found no effect either way.

That last detail matters. Not every AI tool helps. *The important word is designed. The studies showing real gains are not studies of any AI tool dropped into any classroom. They are studies of platforms built on how people learn: tools that adjust in real time to each learner, that demand active retrieval rather than passive review, and that calibrate challenge to keep learners in the productive zone of difficulty

This is the opposite of cognitive overload. This is cognitive load, optimised.

What's Actually Failing Children, and What Would Help

The real failure in education is not too many tools. It is too little thinking about what learning requires for each individual child. A classroom of thirty with one teacher and one curriculum was never capable of the immediate, personalised, calibrated feedback that cognitive science tells us produces lasting knowledge. That was always the fundamental constraint: every child learning at the same pace, receiving the same explanation, assessed at the same moments, regardless of where they actually are.

Mo Gawdat, former Chief Business Officer at Google X and author of Scary Smart, has argued that the question is not whether AI will transform human capability, but what we choose to do with that transformation. His vision is one where technology handles the cognitive scaffolding precisely so that human time is freed for what only humans can give: connection, mentorship, inspiration and emotional intelligence, the relational work of growing up.

Yuval Noah Harari's Nexus draws a harder line: information is not the same as knowledge, and platforms that make information cheap and abundant do not, by themselves, build wisdom. The concern is valid. But the answer it implies is not fewer tools. It is tools that demand wisdom-building, that ask children to reason, retrieve, apply, critique and create, rather than tools that deliver content and reward passive recognition.

What ExelAcademy Believes, and Builds For

01 Cognitive load is the mechanism of learning, not the enemy of it

Managed well, it is how the brain builds lasting knowledge. The goal is to design for the right kind of difficulty, not to eliminate challenge.

02 Retrieval, not reception, builds memory

Any tool, digital or analogue, that only asks children to receive information rather than reconstruct it is not teaching them. It is entertaining them.

03 Personalisation is not a luxury

The child who needs more time on a concept should have it. The child ready to move ahead should not be held back. No teacher managing thirty children can deliver this alone. Well-designed AI, like our ExelTutors™, can.

04 Technology should create more human time, not replace human connection

When the cognitive scaffolding is handled by systems designed for it, teachers, mentors and parents are freed for what matters most.

05 The standard should be the science, not the sentiment

Nostalgia for workbooks is not a pedagogical argument. Neither is fear of technology. The only question worth asking is: what does the evidence say produces deep, durable, transferable learning?

Forward to Fundamentals

The children who will lead the next decades of human history are in classrooms right now. They are being shaped by what we decide education means, and by whether we have the courage to let that decision be driven by evidence rather than anxiety.

Going back to basics is not the answer. Going forward to fundamentals is.

The fundamentals of how the brain actually learns. The fundamentals of what children are genuinely capable of when we stop limiting them to what the nineteenth century could imagine. The fundamentals of what it means to prepare a human being for a world that has never existed before.

That is the work. It has never been more urgent.

Tirzah Elese Bagnulo

About Tirzah

Tirzah Elese Bagnulo is Founder and CEO of ExelAcademy, a human-led, AI-enabled online school. She brings more than three decades of teaching and educational transformation, fifteen years of studying how the brain learns, and is the author of The Learning Commitment. Tirzah Talks is where she thinks out loud about what education is for, now that AI can do so much of what it once focused on.

Sources & further reading (8)+
  1. Fortune. How did US spending $30 billion on laptops result in the first generation less cognitively capable than their parents? (February 2026; Jared Cooney Horvath, US Senate testimony).
  2. Sweller, J. (1988). Cognitive load during problem solving. Cognitive Science; Sweller, van Merriënboer and Paas (1998, 2019) on intrinsic, extraneous and germane load.
  3. Kwak, M. (2025). The effectiveness of AI-driven tools in improving student learning outcomes compared to traditional methods. Issues in Information Systems, 26(4).
  4. Kestin, G. et al. (2025). AI tutoring outperforms in-class active learning: an RCT. Scientific Reports (Harvard University).
  5. du Plooy, E., Casteleijn, D. and Franzsen, D. (2024). Personalized adaptive learning in higher education: a scoping review. Heliyon, 10(21).
  6. Discover Education (2025). Artificial intelligence in adaptive education: a systematic review of techniques for personalized learning (142 studies).
  7. Gawdat, M. Scary Smart (2021).
  8. Harari, Y. N. Nexus (2024).

More Tirzah Talks

Find more blogs in Tirzah's Homeschool Network

Join the community for new Tirzah Talks posts, ready-to-use project-based learning, and a place to share your thoughts with families who believe education can be better.

Join