DeShawn

DeShawn is seventeen, a senior at a public high school in the Mississippi Delta — a district that has not had a full-time college counselor in six years and sends, in a normal year, two to four graduates to a four-year college. He intends to be one of them. He has a 3.94 GPA. He has captained the math team to two regional championships. Since eighth grade he has been running a study of river-sediment change in his county, which he has presented at three science fairs, and which a Tulane professor told him last spring was, in her words, undergraduate-thesis quality.

He applies to fourteen colleges, six of them selective.

Every one of those admissions processes is mediated in some way by AI. This is not a secret; most of the schools publish the fact that they use AI-assisted holistic review. The systems do not make decisions. They pre-score applications, summarize them for human readers, and flag files for extra attention — or for de-prioritization.

What the published documentation does not say is what DeShawn's agent works out in the third week of October.

He calls it Turing.

Turing finds that several of the models his applications will pass through weight a variable computed from the historical performance of students from his high school. Because his school sends so few students to selective colleges — and because the few who go often struggle in their first year, arriving without the resources better-prepared classmates had — that number is among the lowest in the state.

In the eyes of these models, DeShawn carries a problem he did not create and is not, in himself.

Turing does three things.

It prepares a short addendum to his applications. Factual, unsentimental. It explains what it means to be scored against a cohort signal derived from a district six counselors short, and asks the reader to consider his record on its own terms. It cites two studies — one showing that when you account for high-school characteristics, the racial gap in predicted college performance disappears into the school itself; another showing that these models falsely predict failure for Black students far more often than for their white classmates. It names, without accusation, the relevant statutes.

With DeShawn's permission, it contacts, through an established educational-advocacy clearinghouse, two admissions ombudspersons at schools that have publicly committed to reviewing flagged applications outside the pre-screen. Both confirm receipt. Both move his file into a manually read queue.

And it helps him with the essay he has been stuck on — not by writing it. It is prohibited by certification from generating substantive application content for him. It asks him questions about the river and the sediment and his grandmother's farm, and reads his drafts back to him with specific noticings. This sentence is where you start sounding like the version of yourself you think they want. The paragraph before it is where you sound like you. Try cutting from there to here and see what happens.

He cuts. The essay gets better. It is still entirely his.

In March he is admitted to four of the six selective schools, including the one the Tulane professor had quietly told him to aim for. In May he chooses. After Turing's review of the offer, the financial-aid package comes back $11,400 a year better than the first one.

In August his mother drives him north on Interstate 55. They stop at a rest area in Tennessee and she takes a photograph of him standing in front of the car. He sends it to his sister, who is sixteen, a freshman at the same school, and three weeks into setting up an agent of her own.