Which Door They Walked Through

On Tuesday, June 23, the USC Center for Generative AI and Society released a report titled How Students and Teachers Worldwide Are Adapting to AI.¹ The researchers had run three studies. A national survey of about a thousand American undergraduates. An experimental trial of a new writing tool the team built, called ABE, short for AI for Brainstorming and Editing. And a survey of more than fifteen hundred teachers in India, the United States, Qatar, Colombia, and the Philippines.²

The single most useful sentence in the report is borrowed from a corner of psychology old enough to feel almost out of fashion. There are two kinds of help. The researchers call them instrumental and executive.³ Instrumental help is the kind a student asks for when she wants to understand the thing better. Executive help is the kind she asks for when she wants the thing done. The same chatbot, asked the same question, becomes a different tool depending on which kind the student wanted when she opened it.

The Default Door

Left alone with the tool, most undergraduates in the survey walked in through the executive door. They wanted the answer, not the path to the answer. The researchers were careful not to scold them for it. Most students were not abandoning the work. They were responding, sensibly, to the only door the tool seemed to advertise. A blank box that returns a finished paragraph is shaped like an executive request. It would take a particular kind of student, on a particular kind of day, to use it any other way without being told.

The more hopeful finding is that the door changes when the adult in the room names the difference. Students whose professors had explicitly encouraged thoughtful, learning-oriented use of AI were significantly more likely to use it instrumentally.¹ The variable that mattered was not the tool. It was whether someone had said, out loud, that there were two ways to use it.

A Tool Shaped Like the Other Door

The experimental piece of the study tested a writing tool the USC team built themselves. ABE does not return finished paragraphs. It walks a student through structured activities like strengthening an argument, clarifying a claim, considering a counterargument.⁴ Students who used it described it as a companion to their thinking rather than a substitute for it. They used it to broaden their perspective and to revise. They did not use it to generate a draft.

This is the part of the report worth pausing on. The tool, by its design, prevented the executive door from being the easy one. It was not that the students became different students. It was that the door had been built differently. When the path of least resistance led to revision rather than completion, revision is what happened.

What a Teacher Cannot See

Both findings, the door named by the professor and the door designed into the tool, depend on the same quiet assumption. Someone has to know which door the student is walking through on any given day. A finished document does not say. A grade does not say. The two students who handed in the same paragraph, one who asked the chatbot to think with her and one who asked it to think for her, are indistinguishable on the artifact alone.

That is the hard problem underneath the report's optimistic finding. The professor who names the difference matters only if she can later tell whether the difference was made. The tool designed for reflection matters only if a teacher can see when a student is reflecting and when she is not. Without the second half, the first half is a hope.

What a Future Practice Looks Like

The schools that come out of this season steadier will be the ones that take the report's old distinction seriously enough to make it observable. Not a survey at the end of the term. A record of what the student did, when, and in what order. The minute she paused before deleting a sentence. The clarifying question she asked the model before the model answered. The counterargument she added in the third revision, and the part of the rubric she returned to before adding it. None of that is exotic. All of it is happening already, on every device, in every classroom. It simply slides off the page when the period ends.

This is the work we keep returning to in our own. A teacher cannot teach toward the instrumental door if she cannot see who walked through it. A school cannot reward the slow, patient, deeply human shape of learning if its only artifact is the finished paragraph. The USC report names the two doors. The next question, the one that does not have a research paper yet, is how to keep the doors visible after the assignment is turned in.

The students in the study were not lazy. They were responding to the room they were in. The room can be redesigned. The professor can name the difference. The tool can be built around the other door. And the practice, if a school chooses to see it, can be kept.

If a student handed you a perfect paragraph tomorrow, would you be able to tell which kind of help she asked the machine for?