AI & Technology

Keep the Thinker in the Research

Write the question and the doubts before a tool is opened. Then verify.

The risk I worry about is not a polished sentence. A polished sentence can be revised. The risk is a skipped judgment: the moment a student accepts a frame, a source list, or a counterargument they never had to build. Once that happens, the draft can look complete while the thinker has left the room.

I use emerging tools in class. I also keep a bright line around authorship. Tools can help students find, question, and revise. They do not get to own the claim. The AI research protocol is the version students keep in the front of the folder. This is the reasoning behind it.

Think before the tool is open

Students write three things by hand, or in a document they cannot pretend the tool wrote:

  • the question, in one sentence, with the scope visible
  • the kind of evidence that would actually help
  • two doubts they already have about their own hunch

If those boxes are empty, the tool is closed. This feels slow the first week. It is the whole point. A tool is most dangerous when a student does not yet know what they are looking for, because then any fluent answer will do.

I will allow a tool to suggest search terms, to translate a confusing paragraph into plainer language the student must then check, or to propose a counterargument the student has to go verify. I will not accept a tool’s paragraph as the analysis. The difference is whether the student’s judgment is still the thing being practiced.

Verification is the assignment

Whatever a tool offers, a source has to exist outside it. Students retrieve the source, read the relevant part, and record whether the tool represented it fairly. Misrepresentation is a finding. It goes in the process notes. I would rather see a student catch a distorted summary than watch them cite a source they have not opened.

The process note is short on purpose:

  • What did I ask the tool to do?
  • What did I accept, reject, or rewrite?
  • Which source confirmed the part I kept?

If a student cannot answer those, the material does not go in the paper. That rule applies to a search summary, a grammar suggestion that changes meaning, and a classmate’s outline. Convenience is not a citation.

Voice is a test, not a vibe

I ask students to read two paragraphs aloud and mark any sentence they could not explain to the person across the table. The mark is not an accusation. It is a revision plan. Academic writing is allowed to sound formal. It is not allowed to sound like a person who is no longer present.

We talk, too, about privacy. Another student’s draft is not raw material for a public tool. Personal stories from a seminar stay in the room. The classroom norms put that in writing so I am not inventing the rule after something goes wrong.

Why I do not outsource the ban

A total ban is simple to announce and hard to learn from. Students still need a way to tell a useful suggestion from a stolen thesis. They will need that way in college, at work, and in the group chat tonight. My course is a place to practice the distinction while a teacher can still ask, “Where did this reason come from?”

The standard I can say to a parent, an administrator, and a senior in the same words: the thinking in the paper has to be thinking the student can defend in the room. Tools that help them reach that standard can stay. Tools that replace it cannot.