In Defense of an Unpersonalized Education

I spoke before about the unintended side effects of good products. When we only think about the benefits we want to create, we can risk introducing new problems to processes that already work. I believe this has been happening (and continues to happen) in education, in particular around personalization.

Outside of education, personalization has long been regarded as the peak user experience. After all, who doesn't love hearing their own name? Cognitive scientists have long known about the cocktail party effect, first described by Colin Cherry in 1953, the phenomenon where we instantly notice our own name amid background noise. Marketers weaponized this: personalized subject lines, targeted ads, recommendation engines. And since then, we've been trying to personalize experiences as much as possible.

Amazon is a great example of this where it uses your search and behavior history to try to suggest the exact thing you want to buy, and it's creepily good. What we rarely acknowledge is that this level of personalization requires an enormous amount of tracking. The same type of surveillance infrastructure that recommends your next purchase is now going to be needed to personalize learning experiences. (But this post isn't about data privacy.)

Personalization, in a way, is just another way to make things easier, more palatable and more engaging. So that should be great for learning, right? Maybe not.

I want to argue that personalization in EdTech teaches the wrong meta-lesson: that learning should flow easily, be tailored, and frictionless. Three of the most important durable skills - adaptability, resilience, and collaboration are only truly learned from real, meaningful challenge that doesn’t care about your needs, comforts, or interests.

The Genuine Benefits (And The Trap)

Before I get into all the problems of personalization, I want to note that there are several genuine benefits of personalization. I do believe learning should have some degree of scaffolding, and that we can't throw any content at any learner and expect them to figure it out. There's a reason we have thought-out curriculums that build one idea on top of another. Personalization is why tutoring works and why allowing teachers to focus on individual students really supports students' learning. But there's a lot more to that type of personalization that the EdTech flavor of personalization leaves out.

Social personalization via a tutor or a teacher adapts to challenge you (emotionally and intellectually). Algorithmic EdTech personalization adapts to keep content interesting and the student engaged. It's curation, not tutoring.

A tutor challenges you based on what you know, but also who you are. They know how to push. They also care about you, and when you are being challenged by a mentor or teacher, that challenge, even if intellectually demanding, feels safe. This is based on trust, common ground, and the invisible social chemistry of a human relationship. But the EdTech version of personalized learning only focuses on the curation element of what tutors do. Sometimes these systems can curate content aligned with level and interest even better than a tutor. But it misses the overall benefit of social personalization. We often mix these up, which leads to a lot of investment and hope being put into AI personalized learning.

“Imagine a tutor for every child,” EdTechs say, who are definitely not providing a tutor for every child. They’re providing a machine, designed to curate like a tutor. The trap is that machines can do a lot of things better than people, but they will never be people. (No, not even Claude.)

Four Ways Personalization Undermines Learning

1. Friction is the Pedagogy

Learning is about doing hard things. If it comes easy to you, it's unlikely you're learning anything new. One of the biggest challenges with EdTech is that you actually don't want to reduce friction, because friction is where you learn. (One upside is that this has created one of the most interesting sub-disciplines of UX - where usability but not ease is the gold standard.)

Research on desirable difficulties from psychologist Robert Bjork shows that learning is stronger when it feels harder. The experience of having to expend effort, generate errors, or work hard to achieve understanding is not evidence of inadequacy, it's the mechanism. The struggle is the point.

Personalized EdTech removes that friction. It shortens the path, smooths the road, and in doing so, removes the very thing that makes the process of learning work.

2. It Narrows Your World

Think about what if you only learned things that were interesting to you, or you only learned through examples that were relevant to you. How would that prepare you for a world that is often filled with boring problems, where the interesting stuff comes from doing a lot of boring work? By personalizing education, we are actually teaching students a meta-lesson: that the only things worth learning are the things that are interesting to you.

Education researcher E.D. Hirsch has argued that a broad, shared base of knowledge is essential for later learning. In The Knowledge Deficit, he writes: "The more you know, the more you are able to learn." If students only study what algorithms deem "relevant" to them, they skip the foundational, seemingly dull material that makes advanced understanding possible.

But here's the deeper problem: personalization skips connections when it skips content that's not a good fit for the learner. Learning is cumulative, not just in content, but in meaning. If we only learn what's interesting to us, we end up with massive gaps of knowledge, and worse, we never learn why the boring parts mattered.

3. It Kills Adaptability

Perhaps the greatest loss from personalization and engagement-focused EdTech: doing difficult work isn't always rewarding, and resilience through boring things is actually the most important durable skill.

Personalization makes learners less adaptable. If students are learning things that are perfectly tailored to their interests, to their reading level, and the modality they like to learn in, they'll start expecting the rest of the world to also adapt to them. But what we really want to be teaching them is how to adapt themselves to a changing world.

The World Economic Forum's Future of Jobs Report 2025 consistently ranks adaptability, resilience, and flexibility among the most critical durable skills for the future workforce. Yet personalized algorithms can’t teach this because they are designed with the opposite goal in mind. These algorithms (or in some cases AI systems) tell the learner they don't need to be flexible. They don't need to power through anything - we can just add more scaffolding. The system adapts to the learner, training them to be passive recipients of perfectly fitted content rather than active agents who can wrestle with mismatch and discomfort.

This is something personalized education can never teach, and in fact, an unpersonalized education can. What is an unpersonalized education? It's taught at the school of hard knocks. Their school colors are black and blue. There's no diploma, but there's no mistaking someone who's graduated. Graduates just have that figure-it-out-ability and comfort with leaning into problems (and comfort with being wrong at first).

4. It Weakens Collaboration

Personalization and collaboration are structurally incompatible. If every student is on their own "personalized learning journey" at their own pace, in their own modality, consuming content curated to their individual interests, then what exactly are they collaborating on? Collaboration requires shared context, which is another thing that is lost when we lean hard into AI-powered personalized learning.

When everyone learns the same thing at the same time, the bored student helps the struggling student. The student who grasps it quickly explains it to the one who doesn't. Meta-learning happens in discovering how you differ from others, where your strengths complement their weaknesses, how a group becomes smarter than its smartest member. Personalized learning paths eliminate the togetherness that makes this possible. We become better when we work together, and learning together is by its nature not personalized... unless of course we are now in our own classroom filled with AI agent peers that are personalized to be the perfect peers for each of us. (Please, nobody build this.)

The Deeper Philosophy

Productive Struggle

Robert Pondiscio posts an important question in The Illusion of Learning: The Danger of Artificial Intelligence for Education. The question is, well what is learning really about, anyway? Learning always comes down to productive struggle, which Pondiscio describes as the effort of sorting, comparing, rephrasing, and struggling to make sense of things. It's not about acquiring knowledge or skills (that's the outcome, not the process). Learning only happens when we're willing to see that we don't know. When everything is scaffolded, personalized, and adapted, the process of learning might be lost. One thing Bjork points out is that learning is not visible, only performance is. When building products we hyperfixate on performance, which makes the learning process easy to miss and its loss becomes one of those unintended consequences I spoke about in my previous post.

Inconvenience as a Modality

One of the most important hard things personalized education leaves out is inconvenience. Confusion, boredom, even frustration. Are these pedagogical flaws? Or are they actually educational modalities? If we're teaching humans in an environment that maximizes comfort and personalizes every lesson, what happens to them when they're put in a challenging situation? When they're faced with choosing a path no one has carved for them? How can we build resilient, intelligent, thoughtful, brave people if all their preparation for life has been set up in a context where they're not inconvenienced?

The One-Room Schoolhouse

The one-room schoolhouse is an interesting educational model because the teacher couldn't personalize for thirty students across eight grade levels. That impossibility was the feature. Students taught each other. Older kids explained concepts to younger kids, solidifying their own understanding. Everyone overheard lessons not meant for them. The curriculum was shared, the pace was collective, and the discomfort of not being the center of attention was constant. There’s something to be learned from conditions it created: forced collaboration, peer teaching, and the understanding that your education was not designed around you. Perhaps going backwards towards something that looks like this might be the real innovative path.

Life’s Repeating Lesson

Perhaps the most important value learners should take away from their education is that it's not about you, but you are capable of figuring out any problem as long as you put in enough effort. And that's a lesson we've flipped: we tell kids, it is all about you. What are you interested in? What modality do you prefer? Do you want to learn faster, or slower? Do you want a video? Or a quiz? Or a game?

In life, the challenges we face are not personalized for us. These experiences do hold up a mirror to us. They can show us the skills we have available, and if we pay attention, through these challenges we can learn about the skills we lack, and through resilience we can meet these challenges.

In other words, life throws shit at you, and that shit is not personalized.