Critical thinking in the age of AI: notes from a design classroom

AI gives fast results — but learning to design still needs the slow, failure-filled process that builds real judgment.

This piece is adapted from a presentation I give my design students at the start of each course. The classroom stories, ideas, and views are my own; I used an AI writing tool to help draft and edit the piece across several rounds of revision.

What I tell design students on day one of every course

Every semester, before we open a single design tool, I sit my students down for a conversation that has nothing to do with typography, color theory, or prototyping. We talk about thinking.

Specifically, we talk about what happens to our thinking when AI starts doing so much of it for us.

I teach Emotional Design, App Design, and Interaction Design at Shenkar College, and over the past couple of years I’ve watched AI tools move from a novelty in my classroom to the default starting point for almost every project. That shift is not inherently good or bad , but it is a shift, and it deserves to be named before students dive in.

So instead of jumping straight into briefs and sketches, I start with a discussion. I ask the room a set of questions I don’t fully answer myself:

• How are you using AI day to day, or in your studies — and where does it help versus make things harder?
• Do you see AI as a creative tool, or a technical one?
• If everyone is using the same tools — ChatGPT, Midjourney, etc. How do you hold onto a design voice that’s actually yours?
• What part of the design process can never be delegated to a machine?

There are no right answers here. The point is to get students noticing their own relationship with these tools before that relationship calcifies into a habit. When I ask that last question, what part of the process can never be delegated to a machine , the answers tend to cluster in two places: “creativity,” which is usually said with more conviction than definition behind it, and “understanding people,” which is vaguer to name but, in my experience, closer to the actual answer.

What critical thinking actually is

We throw the phrase “critical thinking” around a lot, so it’s worth slowing down on what it actually means.

Critical thinking is the ability to process information, understand context, sit with open questions, form your own opinion, and make a decision. None of that happens automatically. It’s a slow, gradual process built through trial and error — one where you’re forced to gather data, organize your thoughts, and wrestle with contradictions. It’s less a single skill than a muscle, and like any muscle, it only gets stronger through repeated, effortful use.

For designers, this isn’t an abstract philosophical exercise — it’s the job. We’re constantly searching for visual solutions that communicate something to another person. To do that well — visually, conceptually, emotionally , we need context, research, and a real understanding of what we’re trying to say and who we’re saying it to. We also need to understand why a solution sometimes fails (maybe it’s too on-the-nose, maybe it’s too obscure), and we need to know how to take feedback and actually act on it.

Design thinking and critical thinking are the same muscle

Look at the stages of critical thinking side by side with the stages of a design process, and the overlap is almost exact:

Critical thinking: Identify and define the problem → Gather information → Evaluate and analyze that information → Generate possible solutions → Weigh alternative solutions → Make a decision → Implement the solution → Reflect on the process.

Design process: Receive the brief → Understand the audience and context → Gather visual references and research → Sketch → Revise based on feedback → Select and develop a direction → Execute the design.

They’re not two skills. They’re one skill, wearing two outfits.

What happens when AI enters the process

Here’s where it gets interesting — and a little uncomfortable.

When you introduce AI into that same workflow, the process can collapse dramatically: receive the brief → feed the brief to an AI engine → receive a solution. Three steps instead of eight.

That compression can be a genuine gift. AI can deepen our work and speed up parts of the process that used to eat up time without teaching us much. But the same compression can just as easily flatten and shortcut the very process that builds independent thinking in the first place. For students especially, the process is arguably the most valuable part of a project right now — not the final artifact. It’s where you develop research instincts, learn to understand an audience, figure out how to reach the right solution, evaluate alternatives, and give and receive feedback. Skip the process, and you skip the training. (Iteration isn’t just slower — research on iterative design has found it produces measurably better results the more rounds it goes through.)

To be clear, this isn’t an argument against AI. Students today face a genuinely difficult reality: they need to know how to use AI — it saves enormous amounts of time and can genuinely help — and at the same time, learning to design is about trying, failing, learning, and trying again. If they skip that process, or move through it in a handful of giant AI-assisted leaps instead of many small human ones, they can’t see the gap in their own development — the step that would have shown it to them is the one they skipped.

I see this split most clearly at the feedback stage. When I give notes on a piece of work, some students go back and rework it manually — sit with what I said, wrestle with it, make the change themselves. Others take my exact feedback straight back to the AI and ask it to fix the piece accordingly. Both end up with a revised file. Only one of them ends up with a better sense of their own judgment.

This isn’t just a hunch. Recent research on AI use points to something concrete: people who rely on large language models without engaging in their own cognitive processing — recognition, memory, analysis, learning, reasoning — struggle to explain their own decisions afterward. They report a weaker sense of ownership and connection to their own work. MIT Media Lab’s widely discussed 2025 study, “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task” (Kosmyna et al.), found measurably weaker brain connectivity among LLM users compared with those working unaided or with a search engine. It’s a small study and it’s drawn its share of pushback on sample size and methodology, so treat it as a signal worth taking seriously rather than a closed case — but it lines up with what a lot of us are noticing anecdotally in the studio.

In other words: outsourcing the thinking doesn’t just change what you produce. It changes what you’re capable of explaining, defending, and feeling responsible for.

So how do you stay critical while using AI?

I don’t tell my students to avoid AI — that ship has sailed, and pretending otherwise would be doing them a disservice. Instead, I give them a set of guiding questions to ask themselves while they’re using these tools, as a way of keeping their own judgment in the loop.

The source question. Do I know where the data behind this AI output actually came from? Is what I’m getting shaped by cultural, gendered, or commercial bias I haven’t accounted for?

The intent question. Was my prompt actually precise? Did I define context, audience, goals, and constraints as carefully as I would in a normal design process? Does the answer I got actually match what I asked for?

The relevance question. A good idea from AI is not automatically a good idea for the user. Who is this idea actually for? What real need does it solve? Does it account for cultural, emotional, or ethical context?

The purpose question. Am I aware of why I’m reaching for this tool right now? Am I using it to shorten a process, or to deepen a specific line of thinking? Is my goal creation, inspiration, or learning? In my experience this is the hardest question to actually apply, not just understand. A student will often start with a specific idea in mind, ask the AI for help — and somewhere in the back-and-forth, quietly stop pursuing their own idea and drift toward whatever the AI suggested that simply looked good. The original purpose gets replaced without anyone deciding to replace it.

The credibility question. Can I actually trust what I’m looking at? Is the information correct? Is there a gap between what looks credible and what’s actually backed by fact?

The reflection question. What decisions did the AI make on my behalf? Does the solution I got actually answer what I was looking for — or just what was easy to generate? What steps did I skip by using it?

These aren’t meant to be a checklist you run through once. They’re meant to become reflexive — the kind of questions that pop up automatically, the same way a good designer automatically asks “who is this for?” before opening a blank canvas.

Bringing it into the studio

I don’t leave this as a one-off discussion, either. When students start thinking through project ideas, I ask them to bring two ideas they came up with entirely on their own, and two ideas generated by an AI engine. We put them in front of the whole class and pull them apart together: What are the strengths and weaknesses of each set? Were the AI-generated ideas shallow, or actually off-base? How much correction did it take to get the engine to a result we genuinely liked? Do we feel connected to — and do we actually understand — the ideas the AI proposed?

I’ve watched this play out in ways that make the point better than any slide could. One student, given a very specific, narrow assignment, came back with a huge body of work — dozens of genuinely beautiful directions, generated in a fraction of the time it would have taken to sketch even a few by hand. But none of the directions were connected to each other, and he couldn’t make a decision. He’d never gone through the slower process of narrowing, comparing, and committing, so when it came time to choose, he had no internal compass for why one direction should win over another. Another student handed in something similarly outsized: a scope far beyond what I’d actually assigned, full of impressive detail. But the pieces didn’t hold together as a single idea, and there was so much surface material that I could barely find a way into giving useful feedback. In both cases, the volume AI made possible wasn’t the problem — the missing practice of deciding, cutting, and building coherence was.

Before I said a word about either piece, I put them up and asked the class what they noticed. They saw it immediately — the scattered logic, the lack of a single throughline — without me pointing at anything. That’s not because these particular students were behind; it’s because they’re first-years. They can recognize a problem the moment it’s in front of them, but they don’t yet have the instinct to stop themselves from creating it in the first place. That instinct is exactly what the slower version of the process is supposed to build.

We also define, at the very start of a project, exactly where we intend to use AI and why: for creation, for depth, for inspiration, or for learning. Naming the intention up front makes it much harder to drift into using AI as a default shortcut without noticing.

Next year I’m planning something more direct for my third-year students, who already have a design process of their own to compare against: I’ll have them complete the same brief twice — once using only AI, start to finish, and once using only their own thinking and process. Bringing both versions into the room side by side should make the gap between the two approaches impossible to argue around.

The point isn’t to resist AI. It’s to stay awake while using it.

None of this is an argument against AI in design education. It’s an argument for staying deliberate about it. The tools are extraordinary, and they’re not going anywhere. But the habits students build right now — whether they interrogate a source, question their own intent, reflect on what they skipped — are the habits that will determine whether AI becomes an extension of their thinking, or a replacement for it.

I see this across every year I teach — Interactive Design in the first year, App Design and Emotional Design in the third, and mentoring fourth-year students through their graduation projects. The first-years are still fresh enough that it’s genuinely possible to shape how they work before habits set. But exposure to the six questions alone isn’t enough at any stage — and by the third and fourth year, the stakes are higher, because the habits are more set and the work is more their own. If a student is going to use AI at all, they need a specific, trained skill — the ability to actually judge the quality of what the AI hands them, not just accept or reject it on instinct. That judgment doesn’t come from using AI more. It comes from the slower process, the one AI is so good at skipping.

AI can generate a hundred directions in a minute. It can’t ask itself whose need it’s actually solving, or why you reached for it in the first place, or what you quietly skipped to get here. That’s still your job — Asking the source question, the intent question, the relevance question, the rest of them, every single time, until it stops being a checklist and starts being how you think.

I teach Emotional Design, App Design, and Interaction Design at Shenkar College, and work as a product design leader and consultant. This piece is adapted from a presentation I give at the start of each course.

Critical thinking in the age of AI: notes from a design classroom was originally published in Bootcamp on Medium, where people are continuing the conversation by highlighting and responding to this story.