Last week, I asked Claude, an AI model, to summarize a scholarly article. The article was about grief — specifically, the ethics of how institutions communicate with the bereaved. It was a complicated piece, full of hedged claims and uncomfortable silences. The author kept circling back to questions she couldn’t answer, acknowledging limits in her data and sitting visibly with discomfort. There were paragraphs that didn’t resolve, arguments that trailed off into uncertainty. But Claude’s summary came back clean: five sections, a confident thesis and no friction. It read beautifully, but it also missed everything that mattered — the hesitation, the unanswered questions, the texture of someone actually thinking through something hard. The AI had smoothed the whole thing down until it was easy to grasp. This moment made me realize, once more: I no longer trust smoothness.
This semester, I’m teaching a course on memory and AI. The material is inherently difficult: 200,000 hours of Holocaust survivor testimony filled with fragmented memory, ambiguity, silence and pain. We use common AI tools not to write, but to see what remains human in the new age of generative AI. Each week, students run the same testimony through multiple models. The machines summarize, smooth and resolve contradictions that survivors left unresolved. They fill silences that were meant to stay empty. What we’ve discovered is that AI is exceptionally good at producing fluency — and that fluency is often an illusion. Video testimony resists the kind of closure that large language models are trained to provide. And so the course has become a seminar in friction: How to sit with what doesn’t resolve, how to notice what’s been sanded away, how to read for the rough edges that make something real. My students no longer ask whether writing sounds polished; they ask whether it sounds true and loyal to the materials at hand.
I’ve started to notice this instinct everywhere now. An email arrives, perfectly worded, and I find myself rereading it for what's been sanded away. Who decided this? Who is responsible? Smooth language is not neutral. It’s a design choice — and often a moral one. Smoothness diffuses responsibility. Polished statements flatten moral differences. Seamless processes erase who bears the cost. When something sounds too polished, I suspect that something human has been removed: doubt, disagreement, slowness, accountability and the evidence of care.
To be sure, this suspicion didn’t begin with AI, which is just the most visible symptom of our culture that has long rewarded polish over risk. In education, we see this impulse constantly: Rubrics that prize ‘flow’ over intellectual courage, participation grades that reward fluency over hesitant but deep engagement, essays that sound beautiful and say almost nothing. For decades, we’ve built systems that treat friction as failure — something to optimize away, to sand down, to make disappear. AI didn’t invent this culture: it thrives in it. But friction is where thinking actually happens. When everything feels seamless, I start to wonder what we’ve lost.
It’s worth noticing when courses are designed, deliberately, to feel difficult. Take, for instance, GOVT 1109: “Disagreement,” a one-credit course taught by College of Arts and Sciences Dean Peter Loewen. The premise is deceptively simple: disagreements are at the heart of a healthy university and democracy. Each week, students gather to debate contentious questions — capitalism, curriculum requirements, family size, public policy — alongside faculty from different disciplines and visiting voices, including journalists and scholars from other institutions. The goal is not consensus but capacity: Learning to sit in tension, to hold a position while genuinely hearing another, to be changed by an argument without feeling defeated. In other words, the course is built around friction, not fluency. As Hannah Arendt once argued, politics begins in plurality — in the fact that we do not see the world the same way. Smooth consensus can be comforting, but it is rarely democratic.
This distinction between smoothness and friction also points to a larger point. The ancient Greeks understood that harmony does not mean smoothness. Heraclitus compared harmony to a bow and a lyre — forms that hold together precisely because of the tension within them. Education works the same way. It is not meant to feel easy but to make us capable: capable of sitting with disagreement, of not knowing right away, of being wrong in public and revising ourselves afterward and of hearing something unsettling and choosing to stay in the room anyway. AI and institutional habits promise relief from discomfort. They smooth the edges, manage the tension and deliver answers that do not require us to struggle. But relief is not the same as growth. A healthy university is not one where everyone sounds seamless — it is one where people learn to withstand the tension of disagreement and exit changed by the encounter.
What would it mean, then, to treat friction not as failure, but as method? I’ve been thinking about that question in small, ordinary ways. What if an essay were allowed to leave something unresolved? What if a sentence risked being imperfect in order to be honest? What if we designed assignments that explicitly left room for struggle? What if a discussion ended without consensus — and that was understood as progress rather than breakdown? Friction, I believe, is not inefficiency but the very texture of thought. And in an age when smoothness is easier than ever to produce, the willingness to struggle visibly, imperfectly and collectively may be one of the most honest practices we have.
So when something is praised for ‘flowing well,’ I’ve learned to pause. Flow can mean clarity — and clarity matters. But flow can also mean the difficult parts have been polished away, the seams hidden so completely that no one has to feel where the argument resists. I’ve started asking a different question: not whether something sounds right, but whether it leaves room for someone to push back. If it does, there is still life in it. And if it doesn’t, that may be an invitation to press gently against it, to ask the harder question and to keep the conversation alive.
The Sun is interested in publishing a broad and diverse set of content from the Cornell and greater Ithaca community. We want to hear what you have to say about this topic or any of our pieces. Here are some guidelines on how to submit. And here’s our email: opinion-editor@cornellsun.com.
Jan Burzlaff is an Opinion Columnist and a Postdoctoral Associate in the Program for Jewish Studies. Office Hours (Open Door Edition) is his weekly dispatch to the Cornell community — a professor’s reflections on teaching, learning and the small moments that make a campus feel human. Readers can submit thoughts and questions anonymously through the Tip Sheet here. He can also be reached at profjburzlaff@cornellsun.com.
