The Magic Trick Trap: When Scaffolding Becomes Sterility

I have a confession: I am often an accomplice in the sterilization of learning. Recently, while designing an activity for an Adaptive Learning Systems class, I chose the “safe” route. I invented a neat, self-contained case study involving a “magic trick” to illustrate how a cognitive tutor might function.

In the learning sciences, a cognitive tutor is an architecture designed to act like a human mentor: it tracks a student’s mental steps, provides “just-in-time” feedback, and adapts its instruction based on a model of the learner’s mind. It is a complex, high-stakes endeavor. By using a magic trick as the problem space, I made the concept fun and relatable. I provided what I thought was a “soft landing”.

But the moment I finished, I felt that familiar, nagging “ick.” I had provided comfort when I should have been inviting a “stretch”. By polishing the problem until it was shiny and manageable, I was operating under a low-expectation assumption. I was treating my students as consumers of a service rather than epistemic agents—individuals capable of generating knowledge and revising their beliefs through struggle. When we make the problem space too tidy, we strip away the “deliberate difficulties” that allow for deep, long-term understanding.

The Meta-Education Paradox: Solving for the House

There is a profound irony currently haunting our ivory towers. We sit in schools of education—”meta-institutions” whose entire existence is predicated on the science of how people learn. Yet, we are often starving for authentic problems of practice.

I recently sat in a meeting where leadership lamented a lack of “manpower” and “capacity” to produce podcasts, design media, or bridge the research-to-practice gap. Meanwhile, office hours are filled with students desperate for an apprenticeship—hungry for any chance to apply their thinking to a real-world mess.

We aren’t short of problems; we are just protective of them. We argue that students need a “safe space” to practice, but we’ve often confused safety with sterility. This creates a Matthew Effect at scale: a phenomenon where the “rich” (faculty or senior researchers) get to solve the high-stakes, authentic problems, while the “rest” (the students) are kept static with repetitive, procedural tasks. We have professionalized education to the point where the student is a “client” to be shielded from institutional friction, rather than a collaborator who could help solve it.

Beyond the Procedural Trace: Range and the Taylorist Ghost

We have spent decades over-professionalizing our disciplines, often at the expense of the “unobvious connection”. We are still haunted by a Taylorist model of education: an obsession with industriousness and the repetitive tracing of problems that have already been solved and documented. In his work Range, David Epstein argues that while the world tells us to specialize early, the most significant leaps often come from generalists—people who can integrate ideas across domains. He points out that space exploration reached its height not just through pure engineering, but through the intersection of disciplines, such as applying the structural logic of origami to the deployment of solar arrays in space.

Real-world problems do not respect department lines.

• A literacy student might hold the linguistic key to reframing a data-privacy crisis.

• A physics student might bring a new mental model to a social-emotional learning gap.

When we stick to the “made-up” syllabus, we ignore this potential. We act as if operational creativity—the ability to define what a task even is—is something students can’t handle. In the age of AI, “tracing” is perhaps a dead-end skill. If a problem has been solved and documented, an LLM can regurgitate the solution in seconds. The height of creativity is not answering the prompt; it is challenging the rules that confine the task.

The Living Practicum: Everything is a Problem of Practice

I believe the solution lies in returning to the core of what we do: teaching and learning. If we don’t have the capacity to launch a podcast, we shouldn’t hire a consultant; we should create a media literacy class. We invite students to engage in a dialectical dialogue, using narrative structures to communicate science to a public that is currently losing trust in experts.

If we have a shortage of administrative counseling or guidance, why not turn that gap into a problem space? We can invite students to design their own courses, tracing and gathering the most emerging conversations in the field as their primary text. Everything we do in the space of education can—and should—be a problem of practice itself. Turn it into a course. Make it a practicum.

By inviting students to do independent studies that solve institutional needs, we generate true transferability. We encourage them to engage with the scientific method not as an abstract concept, but as a tool for immediate belief revision and conceptual change. This isn’t just about “volunteering”; it is about operating more effectively without over-hiring or expanding “services” that have no pedagogical heart.

Our service is the core learning experience. It is perhaps the time to stop “instructional designing” the life out of our problems and invite the students to come solve them collaboratively. Perhaps to me, this is the heart of education.

Peace,

N.

I’ve been thinking about this deeply. Last semester, in a course focused on AI, learning sciences, and public policy in resource-constrained environments, Professor Seiji Isotani and I conducted a relatively simple but revealing experiment.

The Flying Object Experiment

We designed a series of cognitive tasks with specific constraints and clear objectives. Students were asked to create a flying object that satisfied three performance conditions:

1. Distance: How far can the object travel?

2. Accuracy: How accurately can it reach a specific destination?

3. Dynamic Targeting: How well can it reach a moving object?

To be a bit “sneaky,” we shared online resources with the students on how to make the best flying objects. Predictably, the top search results were almost always paper airplanes. This influenced the teams’ strategies; they immediately gravitated toward traditional aeronautic designs.

Here’s the results of the teams and the winning prize

Defining Creativity Across Domains

How we operationalize and define creativity is often nuanced. A major part of the conversation that scientists seem to agree on is that creativity is often discipline-specific; one can be musically or artistically creative while remaining limited in another domain.

When we break down the task of creating a flying object, it requires three types of creative endeavor:

• Operational Creativity: How do you even define a “flying object”?

• Cognitive Task Mastery: Deeply understanding the requirements of the challenge.

• Psychomotor Coordination: Understanding the physical interaction between the operator and the object.

In this experiment, the only way to creatively test a hypothesis was through iteration—throwing things and testing them repeatedly. Interestingly, the winning team wasn’t the one that tried to “perfect” a single design. They made dozens of versions and realized that traditional models weren’t working. They discovered that our understanding of a “flying object” is often confined to what language models or Google searches suggest: a paper airplane. They practiced throwing the models and adapted in the time given.

The “Chair” Problem: Moving Toward the Edge

This reminds me of my work with Tina Grotzer on “thinking on the edge.” If I ask you to draw a chair, you will likely imagine a flat surface with legs—something sturdy and elevated, like this one that I asked Gemini to generate

But let’s also look at these examples:

Is this also a chair?

Are these also chairs?

The issue with current AI models is that they often “regress” to the mean; they regurgitate the most commonly accepted concept of what constitutes a “chair.” However, if you operationalize a chair as “whatever you can sit on a surface,” the definition becomes generative.

This is a power that AI has yet to fully seize from humans: the ability to constantly assimilate and epistemologically accommodate new concepts, revising definitions rather than just coming out of an existing dataset.

Challenging the Bounds

So, where should we sit conceptually when we think about education and creativity?

We often stick to a linear model of what constitutes learning. But real creativity happens when we move further away from the center and challenge existing boundaries. “Creativity on the edge” is about the person who challenges the premise of the task itself.

During our test in the class, students complained about the inconsistency of the “moving targets” or the distance measurements. In reality, Seiji and I never set strict rules for those parameters. A truly creative student might ask: “Can I control the basket?” or “Can I go out there and change how we measure the travel?”

I consider that the height of creativity—not just answering a specific set of made-up rules, but challenging the very rules that confine the cognitive task.

A New Framework for Learning

I’m not advocating for rule-breaking for its own sake. There are essential rules of ethics and morality that we must not cross. But we must critically ask ourselves: What rules of learning and education have we been abiding by for so long that they now hinder our creativity?

Perhaps this is what Sir Ken Robinson meant when he asked if our schools kill creativity. Is it time for us to re-imagine what creativity is, and can AI be the tool that finally pushes us toward those new boundaries?

Peace,

N.

[Bản dịch tiếng Việt ở đây]

“Mom, if back then I hadn’t gone into Economics and Business Administration, but instead gone deeper into Math or Computer Science, I’d probably be much richer now, and things would be very different, right?”

“That’s true. Why did you choose Economics & Business back then anyway?”

“Well, at the time I saw people working in Economics and Banking getting big bonuses, so they were rich.”

“So what are you doing now again? You’ve told me a few times, but I still don’t really get it.”

“I collect data, clean data, then write code and software to run quantitative and predictive models…”

[She interrupted me] “Oh, so like the IT guys at my office back in the day?”

“Yes, kind of, partly, but the data are about people, behavior, multimodal data, affective data.”

“Then why didn’t you study IT back then?”

“No, Mom. I don’t do IT. IT is Information Technology, it’s… I do something else, I need AI, computation…” —and then a series of English words that I can’t pronounce very well.

“But IT is processing information and designing technological solutions, isn’t that what you’re doing?”

“Yes, but that’s not complete. If I had only studied IT, I wouldn’t be able to do what I’m doing now.”

“So you studied Business Administration, dabbled in Teaching Physics and English, then a Master’s in Education, and now you can do this?”

“…You’re right, Mom. I’m doing IT.”

“So does IT-ing pay enough to live? Do you make a lot of money?

…

(My salary is just enough to live on, and to save a little.)

That night, I couldn’t sleep. My mother had unintentionally exposed how I myself understand and value knowledge work through simplified frames. I tried to use the language of “experts,” and that vocabulary collapsed almost immediately outside expert and academic communities. Money and livelihood became the most intuitive metric for evaluating the value of a profession. I probably didn’t lose sleep because my mother misunderstood me, but because I myself couldn’t fully explain the nature of what I do. I study the Learning Sciences, and yet I couldn’t make my own mother understand what I do. I regressed to the closest approximating concept: IT.

Epistemology: Science, Education, and the Public

For a very long time, educators and scientists, for many reasons—largely driven by the model of elite research universities and global economic shifts—have been absorbed in doing research while neglecting a crucial responsibility: engaging in dialogue with the public and making their work understandable and visible. One consequence of this has been credentialism, and the widening gap between scientific work and practice, between scientific discourse and public trust in scientific evidence. Unintentionally, this has created a major barrier to knowledge access between the scientific community and the public, prompting the public to question the role of scientists and the real impact of research on everyday life—from vaccine safety, to genetic research, to the adoption and use of artificial intelligence.

For years, people working in educational technology and learning design have wanted to close this access gap. Most platforms and training modules answer one main question: what to learn (90%), and how to learn—mostly in terms of convenience, fun, and engagement (10%). The 90/10 split is my own assumption; I have no formal evidence for it, only observation and experience. Many of the personal statements I’ve helped edit for students applying to study education abroad share a common aspiration: increasing access, going abroad to “save the country,” bringing knowledge and civilization back home :)

And when many international students return, they quickly find themselves in conversations that look exactly like the phone call between my mother and me—struggling to push the knowledge they’ve accumulated into real change and improvement in educational systems. But after prolonged training in scientific reasoning, many of us lose the language and ability to argue with the public. My mother is 55 years old. She studied Russian language, Law, and Political Science. When I first tried to explain quantum physics to her, she used her own language to model and hypothesize quantum entanglement: “It’s like if I’m in Hanoi and I you in Boston, and if you sneeze, I would know immediately, right?” (Not correct, but from that example, I could continue the conversation with her from almost any field.) My mother is a lifelong learner with a special ability to systematize hypotheses (situated and theorized learning) and solve problems—an ability I’m lucky to have inherited maybe one-tenth of. She doesn’t understand STEM—this is obvious—but she has a strong grasp of epistemology and of how I think and reason.

My mother is the most concrete example—and for me, because she is my mother, the perfect example—of an epistemic agent.

Cognitive Science and the Science of Learning in the Age of AI

Four years ago, when I first stepped into cognitive science and the learning sciences, Piaget felt like a monument. He argued that learners are never blank slates; knowledge is constructed through assimilation and organization of experience, new observations, and interaction. But like other foundational thinkers in developmental psychology and education such as Dewey and Freire, Piaget lacked observations of modern technology. Today, learners are not blank slates, but pages densely scribbled with prior information, selective cognition, and behavioral patterns from the past. The challenge for learning design is no longer just content delivery or engagement, but belief revision and conceptual change.

At a macro level, this phenomenon requires educators and instructional designers to return to foundations in psychology and sociology. This is no longer solely a cognitive process problem, but one of socio-cultural identity development and behavioral science. A daily user of AI or LLMs can easily access information that is incomplete, hallucinatory, or even pseudo-scientific, and then use that information to challenge scientific evidence and reasoning. At scale, this creates an epistemic crisis for learners and scientists alike, burying highly reliable, rigorously tested research beneath waves of conflicting information.

Learning Design, AI, and Technology

Educational technologies from 2010–2020 emphasized personalization, engagement, convenience, and access. They contributed to infrastructure and system building, but most only reached the lowest levels of Bloom’s Taxonomy: Remember, and a very limited form of Understand (not deep understanding), alongside factual knowledge. There is little evidence that these technologies effectively support higher-order thinking such as Apply, Analyze, Synthesize, and Create, and almost no evidence for Conceptual, Procedural, or Metacognitive knowledge. Duolingo streaks are a clear example: learners may master vocabulary and grammar structures, yet remain unable to communicate effectively in everyday situations.

When edtech designers focus on personalization, they implicitly make two assumptions:

(1) that personalized content can be delivered with near-perfect accuracy, and

(2) that user design is equivalent to learning design.

The first assumption effectively declares that everyone learns and thinks in the same way, leading to scientifically unsupported ideas such as learning styles, and to systemic designs like tracking, streaming, and age-based grouping from K-12. From Piaget to Vygotsky to Howard Gardner, scholars have pushed back against this ideology, yet educators and investors—from Thomas Edison and Thorndike to Bill Gates and Sal Khan—have repeatedly chosen efficiency and system-level scalability instead. In my view, this is why education has changed so little over the past 100 years, from the United States to Vietnam.

The second assumption is more contemporary and fashionable: “learning should feel like play,” “teaching should be inspirational,” “teaching should be done with heart.”

Yes—but not enough.

When people conflate user design and user experience, which originate from marketing, with learning design, interaction and engagement become ends in themselves. UX should come after learning design. The primary goal must be conceptual change and long-term understanding, and learning necessarily involves struggle and friction. Learning metrics should not be clicks or satisfaction surveys, but whether learners can transfer knowledge to new situations or change their behavior.

At the beginning of the semester, Seiji and I sit down for long conversations while designing a course. How do we make students struggle, ache, and wrestle—without making it so hard they give up, or so easy they get bored? How do we create classrooms where students ask difficult questions, argue fiercely, and critique deeply without destroying peer relationships or trust between students and instructors? This is the hardest and most time-consuming part of learning design—and it is its core.

In the end, the problem is not that my mother cannot understand what I do, nor that I cannot explain it, but that we have collectively normalized a system in which knowledge only circulates comfortably among experts. When complex work collapses into “IT” in everyday conversation, it is not a failure of intelligence, but a failure of translation built into our institutions. In the age of AI, learning design can no longer be reduced to efficiency, engagement, or personalization. Its responsibility is epistemic: to help people reason across domains, revise beliefs, and hold complexity without flattening it. If our educational and technological systems cannot sustain meaningful dialogue between lifelong learners and researchers, then the problem is perhaps not comprehension, but design.

The first time I had a conversation with Yasmin Kafai, she keeps reminding me that AI is a design-technology…

I returned to Longfellow Hall 319–320 after Spring Break in 2022, and an “innovation” shocked me to my core.

As a student who’d studied in that room the year before, I had always loved the classroom. It was a floor-to-ceiling, prize-winning space where students could walk right up to the wall and start scribbling. Any pedagogist or Montessorian would tell you a hundred reasons why that’s such a fantastic idea. The thought that went into the design was evident to every professor who taught there. I would’ve gladly traded any other Teaching Fellow just to teach the 9AM section because of what the space made possible.

And most students at HGSE knew. That room was always in high demand because it offered something rare: the freedom to think out loud, to sketch, to connect ideas, to move. It was designed for learning, not just instruction.

In the Spring of 2023, I was serving as a teaching fellow for Dr. Hadas Eidelman’s popular statistics course, S030. We loved that classroom. The pedagogy we had in mind: asking students to map out their charts, draw regressions, build conceptual models: relied on the 360-degree writable walls. You could draw an arrow, pause, reflect, shift perspective. The room itself became part of the conversation.

But when we came back from break, we found a “remarkable improvement.” The walls had been painted over. Green paint. Fixed, framed whiteboards hung here and there. Small. Confined. And with that change, something essential was lost.

Even at one of the most well-known learning organizations in the world: the Harvard Graduate School of Education, there can still be a disconnect between what the learning sciences know and what institutions actually do. I found myself asking: What happened? Who made the decision? Were educators consulted? And more importantly, how can we build schools that make decisions in service of learning?

Eventually, I asked around. The change had been driven by a real concern: transition time. The white-painted writable walls, while wonderful for learning, were hard to clean. Their rough surface made it difficult to wipe between sessions. Facilities staff couldn’t keep up with the 15-minute turnover between classes. So, in the name of efficiency, new whiteboards were installed.

On paper, that makes sense. I don’t want anyone’s job made harder. I have deep respect for the operations team especially folks like Thomas and JR, who’ve kept this place running through thick and thin. But in practice, it still wasn’t working. I often walked into class only to see the boards still covered in writing from the previous session. The transition time hadn’t gotten shorter. What we lost in the process was far greater than what we gained.

Original link to HILT Site: https://hilt.harvard.edu/teaching-learning-resources/learning-spaces-week-at-harvard/

But it was done. The boards were there. And maybe that’s the heart of it: it’s not about the walls. It’s about what the story of those walls represents.

2. How Learning Spaces Reflect Institutional Priorities

This story mirrors what I’ve seen again and again in education, especially in EdTech: Well-meaning decisions made in the name of efficiency that quietly undermine pedagogy.

It always starts with a real problem. A practical fix. A good intention. But the implementation often leaves behind something essential. And once the change is made, we’re told to make do. “The tech’s already there. Just work with it.”

Sound familiar?

The counter-argument is familiar too: all classes deserve equal treatment, and we must respect scheduling constraints. More efficiency means more students, more access, more learning, more time, more content.

But, well No. We don’t

We need better learning, better time spent, better content: More space for pedagogy to breathe.

3. Human-Centered Design but Who-man?

Two years have passed since those boards were installed, and I still hear debates about human-centered design (HCD). But the version of HCD we practice often only includes a narrow group of stakeholders: faculty, staff, administration. The actual humans the students are often left out.

Designing a classroom without instructors input is like designing a playground without talking to children. Even when instructors are consulted, they’re often the most vocal or visible those in student government or formal leadership roles

The surface-level explanation for the whiteboard change was aesthetic and operational. But beneath it, I see a more troubling trend: the growing belief that education is a service to be consumed. I’ve heard it many times: “I pay, so I deserve this.” Learning becomes a transaction. Instructors become content deliverers. Students become clients. And slowly, the classroom stops being a space for wonder, play, or ownership.

4. The Cost of Treating Education as a Service

I’ve long resisted this mindset. Because when we reduce education to service delivery, we strip away its heart. The unpredictable, co-constructed, joyful mess of learning becomes sterile. Safe. Efficient.

But not alive.

A Small Ritual That Made a Big Difference

One of the simplest, most beautiful practices I’ve adopted? At the end of class, every student grabs a wipe. One per person. In thirty seconds, all the boards are sparkling clean. It’s efficient, sure but more than that, it’s communal. I have never even asked my adult-graduate-level students to do it, not even once. It just happened. And over time, it’s become a quiet ritual that says: This is our space. You belong here. You help take care of it.

That’s what dignified contribution looks like. That’s what real participation feels like in a learning space.

What Does This Have to Do with EdTech?

Everything.

Because too many EdTech products are designed in the same spirit that painted over the white walls. They prioritize ready-launch product over messy and contributive intimacy. Efficiency over pedagogy. Control over co-creation.

Whenever an EdTech founder shows me their latest prototype, instead of asking what it teaches and talking about how personalized or adaptive the design is. I ask:

• Where do students get to choose? How do they get to choose?

• How are they going to make mistakes on this?

• Can they move the metaphorical chairs? Touch the digital walls? Leave a mark?

Because that’s where ownership lives.

The real tragedy of the white walls wasn’t the green paint, or the white boards. It’s that their pedagogical potential was never fully seen: Instead of, “Here’s a room we built for you,” maybe it’s simply, “We’re still figuring it out, want to help?”

When students are invited to contribute, even in small ways, the classroom transforms. It stops being a delivery zone and becomes a shared project. So yes, I’ll keep talking about EdTech design and I’ll keep telling this story. Because these moments remind me what learning spaces are truly made of not just whiteboards and tech, but an active participation in their co-creation of knowledge.

Peace ✌️

N.

I can’t help but trace myself back to the original lecture I first watched from Michael Sandel’s famous series on Justice, which I first saw when I entered high school. It was amazing how far we have come from the ADSL-run internet that took forever to load a 45-minute lesson with very intentional production value, to now getting frustrated when our 4K-resolution streaming service buffers for a few seconds. I got a chance to sit in Sandel’s class last fall in person. The seemingly uninterrupted live service brought directly to my eyes in Sanders Theater. A message remains: in order to learn, to reason with ourselves, to be an effective leader, to just… be a human being—there must be a dialogue.

So here I am, trying to have a dialogue, at first with myself, perhaps as an escape mechanism from all the noise in the world, and perhaps to revert back to the most scientifically proven method of metacognitive development ever known to mankind: writing.

I wanted to start this first note with authenticity in recent times with something seemingly inherent to our nature. I have a nephew, and as I am writing this article, he is just shy of 100 days old. As an educator, I keep asking myself what kind of world we want our children to grow up in. My nephew is cute and adorable, but he is also… and I don’t use this word lightly… just plain stupid. And please bear with me, as there is power to that stupidity. Why? Because he’s learning constantly, but he is yet to be educated. He couldn’t speak or quite fully understand the language, and he is, on the face of it, seemingly authentic and pure.

But is he? By three months, he would probably have already mastered his ability to manipulate his parents’ attention and his grandmother’s care by simply crying. Dare we call him authentic? Dare we deny his need for attention as something of nature, or just simply a “hey, are you there? Look at me, please?” He is perhaps, in the voice of Alison Gopnik, already mastering political philosophy without ever uttering his first word.

The reason I’ve brought all of these stories about my nephew here is because I do believe we are now in a crisis of authenticity. And perhaps for far too long, we’ve blamed it on social media, on the moral limits of markets, on “the way things are.” But perhaps there is an underlying nature to that authentic voice: a dying need to reshape ourselves in the search for something even more important: human connection.

As many parts of us now start to defer these decisions to AI, I wonder if this could be the beginning of a dialogue that Sandel aspired.

[Proofread by ChatGPT, voice still by me… at least for now]

Subscribe now