Taste Collapse: Keeping the Ear for Your Own Voice
A student of mine turned in the cleanest essay she had ever written, and I could not find her in it. The sentences were tidy. The transitions behaved. And the particular crooked way she had of arriving at an idea — the thing that made her writing hers — was gone. When I asked her about it, she did not quite know what I meant. That is the part that stayed with me: not that the voice had thinned, but that she could no longer hear that it had.
There is a phrase making the rounds for this — taste collapse. The idea is simple and unsettling. When you let a machine generate your voice, you slowly lose the ear that tells you whether the result still sounds like you. The skill that judges the writing and the skill that makes it are the same skill, and it is built the same way — by use. Stop using it and it does not hold steady. It fades.
We tend to speak of voice as something a writer has, waiting to be expressed. It is closer to something a writer becomes, through the slow labor of choosing a word, hearing it land wrong, and choosing again. That labor is not the inconvenient part of writing on the way to the product. It is the part that grows the writer. Hand it away and you do not merely outsource the sentences — you arrest the faculty that would have known whether they were any good.
This is not an argument against the tool. It is an argument about which work to keep. Use AI to listen — to gather what you cannot hold in your head, to find the pattern across a hundred student reflections, to draft the scaffolding you will tear down anyway. Then write the thing yourself. Let the machine see what you cannot see. Do not let it speak in your place.
For those of us who teach, the obligation sharpens. Our students will write alongside these systems for the rest of their lives, and pretending otherwise serves no one. What we owe them is not protection from the tool but the development of the ear — the taste — that lets them use it without dissolving into it. That means assignments that reward developed judgment over a polished surface, and that make the labor of revision visible again, precisely because the surface is now free.
The risk was never that the machine would write badly. It writes well enough, more and more of the time. The risk is quieter — that we will write with it until we can no longer tell our own voice from the one it offers us, and will not miss what we have lost, because the instrument that would notice is the very thing that has gone silent. The work worth keeping is the work that keeps the ear. The rest, let the machine carry.