Silence is Golden: Field Notes on a classroom

I had two hours with a group at the very beginning of their learning careers — students who had chosen to study human-computer interaction, which means they had already decided, at some level, that the relationship between people and machines was worth taking seriously. My ambition was not to teach them a tool or survey a field. It was to give them a frame they might carry for the rest of their professional lives: that between any point A and any point B, only two things remain irreducibly human. The first is the question of what needs to happen here — the act of identifying that a problem is worth solving and understanding it with enough precision to act. The second is yes, let's make this happen — the commitment that converts intention into motion. Everything in between those two moments is, to an increasing degree, territory that AI can cover.

This is not a comfortable claim, and I did not present it as one. It implies that a significant portion of what professional education has traditionally trained for — the synthesis, the ideation, the prototyping, the iteration — is undergoing a structural compression. The intermediate steps still matter; they now simply move faster and require less of the practitioner's cognitive labour. What that leaves exposed, paradoxically, is the quality of the two moments that bracket the process: the clarity of the problem definition and the seriousness of the commitment to act on it.

---

Silence is golden

There is a particular quality to silence in a Finnish lecture hall that you don't encounter elsewhere. It isn't the silence of disengagement (waiting for the speaker to wrap up and leave); it is more deliberate than that, perhaps stemming from the very intentional attitude that Finnish culture has around personal space. We can travel along, and we can be silent together, and that is, in itself, the interaction.

When I asked the HCI students at the University of Vaasa to tell me what do you want?, I was purposely throwing in a big question. Nobody knows exactly what they want to do, or who they want to be, much less in your first year at an University. But that answer (or lack of) is what helped me to show how an LLM can help us to get unstuck, and how it doesn’t need to be always precise. In fact, precision would kill the role of inspiration that AI can provide, because inspiration can never be, by definition, a single-answer reply, or it becomes directive.

I prompted my custom GPT, which already works with my Service Design framework, to work with me in this context.

As the ice was breaking from my candid display of discomfort, GPT gave us something to work with:

I then probed the classroom about the first group. Interesting? Hands up. Group 2? So-so. Third group? Okay, I see many hands up here.

When probing the items of Group 3, the class nearly unanimously raised their hands for the Habit tracker specifically for students.

It’s worth noting: we have a specific consumer group (students at a University, n≈40, excluding the online audience) deliberately choosing a new service from over 15 service ideas. Preparation and discussion time ≈15m.

Probing, not asking

I then proceeded with my usual Service Design framework of probing needs, pains, gains and opportunities. This was a delicate ground, because it touched sensitive questions about wellbeing, attitude towards failure and success and self discipline.

But to my surprise, the room was warm and it felt safe to raise hands, vote individually, and even pick between close concepts, considering trade-offs.

A great deal of self-disclosure was required from the room, which responded actively to which need felt more or less close to home.

“Does this sound familiar?” “Do you identify with this?” “Anybody sees this as relevant?”

Those questions were the compass I took to assess if the LLM was getting closer or further from the students’ real needs.

And the students nearly unanimously gravitated towards a habit tracker of their own study habits that would not make them feel worse when they stopped using them. That was all I needed to generate a provocation and a discussion piece, and advance significantly in a demonstration of product creation.

---

Boiling down to the features

Probing proved to be a productive methodology. We got to what they needed from a study habit tracker, what existing methods got wrong, what gains they actually wanted. They were remarkably specific — not because I prompted them toward specificity, but because the exercise asked them to speak from their own lives. The pains they identified converged on something that formal habit-tracking literature has begun to address: that systems built around streaks and daily check-ins inadvertently encode guilt as a mechanism. Miss a day and the app registers failure; the emotional cost of returning rises with each absence until the user stops returning at all (Fogg 2019). The students didn't need to know that research to identify it. They had lived it.

From that diagnosis, we built. Using Claude, and working through a structured opportunity framing they had constructed themselves, the groups prototyped a study tracker whose central design principle was recovery rather than consistency. The application would not punish a student for drifting from their schedule; it would reward the act of getting back on track, treating return as the meaningful event rather than continuity. Within ninety minutes of arriving at that brief, there was a working prototype on screen.

Screen Recording 2026-03-31 at 15.52.43.mov

I distributed book copies to students who asked for them after the session — several did, which in a Finnish classroom counts as enthusiasm. The professor showed me the robotics and AI projects running in the department, including work that would not look out of place in a research lab with considerably more resources. The campus has the calm purposefulness of an institution that has been building something for a long time without needing to announce it.

What stayed with me on the train back was not the prototype, which was genuinely good, but the moment when a room full of students who had mapped a real problem looked at what they had produced and understood that little gap between identifying something worth building and having a first version of it. When I compare this to my first years of design thinking (hand-drawn screens, arrows everywhere, walls covered in sticky notes) I cannot help thinking that’s how the evolution of poster design happened from paintbrushes to computer-mediated. Any romantic attempt to send us back to posters designed with paint and paper falls short, because there was a point that the artists themselves decided the computer can give them more agency over their work. We may be in this change for service and product design as well.

In this scenario (a big room, and a quiet but active participation) the technical work (which was most of the work!) had become nearly invisible. We wasted no time in describing details, researching the best practices for calendar display, or planning screen by screen what could render a valuable experience. We thought about the problem, we curated the opportunities and we drafted a solution, in less than 60 minutes (the second half of the session). The prototype itself did not matter (this can be refined ad infinitum). What mattered was they had asked the right question in the first place, and agreed about a problem that, unanimously for the forty students, was worth thinking about.

Post lecture, and after a guided tour through the campus with Prof. Rebekah Rousi, who I met all the way back then during my Master’s.

Compression and decompression

The current dispute in discussing AI use for creative processes is the loss of the process. I won’t concede that there is no loss, but I think there is a qualitative displacement of the slow processes.

Take, for example, this anecdotal case of the habit tracker for students. We did not waste time discussing how calendars should be displayed (often one of those cases in which UX designers reinvent the wheel and waste precious time trying to figure out how to do it). But instead, we postpone the fine-tuning of the calendar feature, focusing on the big picture first with a very crisp view of what it could look like even in a possible first launch. The fidelity of the prototypes point also to the distance from a first launch, period. We went from “aspirational and good intentions as a paper prototype” to “good enough for launch”, taking only further steps with Claude Code to make it live.

This compression (for good or bad) is the defining condition of the work ahead of them. They will need to deal with this speed, discover the trade-offs, and displace the fast and the slow parts of the creative process.

And before and after creative processes, we have decisions. LLMs are not, at present, good at making them. The students who learn to embrace their role of making decisions and become precise about what they want before they reach for a tool, may find that the tools can serve them very well.

The ones who skip past those two moments in the rush to generate output will produce a great deal of work that answers no question anyone actually asked.

The Finnish silence, it turns out, is excellent training for this. To sit with a problem long enough before speaking, in the age of instant generation, is a professional competence.