Between Relief and Displacement

On May 5, the University of Washington released a study that may be the most honest reading we have so far of how teachers are actually living with AI.¹ A research team led by doctoral student Aayushi Dangol sat down for semi-structured interviews with 22 teachers in Aurora Public Schools in Colorado, a district that has invested heavily in Google's Gemini and MagicSchool.¹ The paper does not arrive at a verdict. It arrives at a name for the tension the teachers keep encountering.

The name is relief or displacement.²

The teachers welcomed what AI could subtract from their week. The duplicate emails. The seventh draft of a vocabulary list. The discussion question for a unit they have already taught fourteen times. They also worried about what was being subtracted. The intellectual work of designing a lesson. The relational work of reading a student's writing line by line and noticing the place where the thinking stalled. The slow accumulation of judgment that made them, over years, the teachers their students remembered.¹

They were not refusers. They were the early adopters. And what they were describing was not a binary, but a pressure they were still negotiating in real time.²

What the Teachers Were Watching For

Read carefully, the study is a careful taxonomy of where teachers feel AI helps and where they feel it begins to hollow them out. Generative AI helped with rote planning, multilingual support, scaffolding for students with IEPs, and the long tail of administrative friction that drains the week. The teachers liked all of that.²

What concerned them was the kind of substitution that does not register at first as substitution. The lesson plan AI produced was usable, but not theirs. The feedback AI generated for student writing was thorough, but not theirs either. Over enough weeks, they worried, the small acts of intellectual ownership that constitute a teaching practice would be quietly outsourced. What would remain would be a teacher who delivered AI-generated instruction to students who submitted AI-generated work. The relationship at the center of the classroom would still be there, but thinned.²

The researchers call this the tension between relief and displacement. It is the right phrase. Relief is real. Displacement is also real. The question is whether they have to come together.

The Future Where They Do Not

It is possible to imagine a version of school where they do not. The condition is simple to state and difficult to build: AI must reduce the labor of teaching without reducing the visibility of learning.

Most current tools do the first and ignore the second. They speed up planning, summarize student work, generate quiz items, draft parent emails. They take pressure off the teacher's calendar. They do nothing in particular to deepen the teacher's view of any individual student's mind. The relief is local. The displacement is structural.

The future the Aurora teachers seem to be reaching for is one in which the time AI gives back is reinvested in the part of the job that AI cannot do. Reading a student's revisions across three drafts and noticing the moment a vague paragraph sharpened. Sitting next to the student who paused for eleven minutes on the third question and asking what stopped them. Writing a comment in the margin that no model could have written, because it depends on knowing this student, this week, in this class.

For that reinvestment to be possible, the act of learning has to be legible. Not a final score, not a finished essay, but the actual texture of the work as it unfolded.

Where Koan Sits in This

This is the bet underneath Aidan, our AI tutor. Aidan does the parts of the work that should be lifted: drafting practice problems, summarizing common patterns of confusion, surfacing which students are stuck and where. It also captures the parts of the work that have always been hardest to see. Every revision a student made. Every pause that lasted longer than thirty seconds. Every breakthrough where a sentence finally landed.

The teacher who opens Koan in the morning does not get a productivity dashboard. They get a record of how their students thought yesterday. The relief shows up as time. The displacement does not show up at all, because the relational core of teaching has been amplified rather than abstracted.

The Aurora teachers are not asking for less AI. They are asking, with the precision of people who have used these tools long enough to know their grain, for a different shape of it. One that takes the rote work without taking the work that made them choose this profession in the first place.

The Question They Leave Us With

The study ends with the observation that teachers are not simply adopting or resisting AI. They are negotiating it, day by day, against their professional identities.² That negotiation is happening right now in tens of thousands of classrooms, mostly in private, mostly without research support.

If the next generation of school AI is shaped by anyone, it ought to be shaped by them.

What would you build differently if you had to design a tool that gave teachers back their week without giving back any of the seeing?