AI Policy Questions from Students

"Can I use ChatGPT to debug this?" is now one of the most common questions a TA hears, along with its cousins: can I paste in the error message, use Copilot, check my grammar, study with it? The answer depends on the course, and courses differ a lot, sometimes between sections of the same course. Your job is not to decide the policy or to share your own view of AI. It is to know your course's policy, explain it the same way every other TA does, send the gray areas to the instructor, and report suspected misuse without accusing anyone. This page shows how, with model answers to the questions you will get.

Defaults that work. If you only take four things from this page:

Know the policy in writing before students ask: the syllabus, each assignment, and the instructor.
Answer from the policy, in the course's words. When it doesn't cover the question, say "I'll check with Professor ___" and don't invent a rule.
Post the answer to gray-area questions for the whole class, so every student gets the same answer.
Never accuse and never rely on an AI detector. Write down what you saw and tell the instructor.

Find the policy first

Look in three places: the syllabus, each assignment's instructions (which can override the syllabus), and the instructor.
Ask the instructor for anything already decided: "Is debugging with AI allowed? Grammar tools? Copilot?" Share the answers with all the TAs.
Agree as a staff on answers to the common questions, so every section hears the same thing.

Policies usually fall into three kinds, and the same course can use different ones for different assignments:

Not allowed: AI tools may not be used on the work.
Allowed for some uses or some assignments: for example, explaining error messages but not writing code.
Allowed with disclosure: students use AI and say how.

At UMass

The default is no. The Academic Integrity Policy counts the use of generative AI as cheating unless the instructor explicitly allows it, and presenting AI output as your own work as plagiarism. Campus guidance tells students that without explicit permission they should assume AI is not allowed, and check with the instructor.
Instructors are expected to state the policy in the syllabus or the assignment.
TAs report, they don't decide. Graduate TAs report suspected dishonesty to the instructor of record, who contacts the student and decides what happens.
Disclosure: the Center for Teaching & Learning suggests students name the tool, say what they used it for, and say what they did with the output.
Student data: never paste student work, names or grades into a free public AI tool. UMass lists which licensed tools are cleared for which kinds of data; even then, use one only if your instructor agrees.

Questions you'll hear

Student asks	If AI is allowed with disclosure	If AI is not allowed
Can I use AI to debug my code?	"Yes, if this assignment allows it. Say in your disclosure what tool you used and what you changed."	"No. Bring the bug to office hours and we'll trace it together."
Can I paste in an error message and ask what it means?	"Usually yes. Disclose it, and make sure you can explain the fix yourself."	"No. Check the documentation or post the error on the course forum."
Can I check my grammar with AI?	"Check the policy's wording. If it's allowed, disclose it."	"The policy doesn't mention grammar tools, so I'll check with Professor ___. Until then, don't."
Is Copilot autocomplete OK?	"Only if the assignment allows generated code. Disclose it if you use it."	"No. Turn it off for this course."
Can I study for the exam with AI?	"Yes, but the exam is closed to AI, so make sure you can do the work on your own."	"The policy is about graded work. I'll confirm with Professor ___ whether studying counts."
Can AI explain a concept from lecture?	"Yes. Check what it says against the course notes; it is sometimes confidently wrong."	"I'll check with Professor ___. Office hours and the forum are always open for that."
How do I disclose?	"Name the tool, say what you used it for, and say what you did with the output."
My friend's course allows it. Why not ours?	"Each course sets its own rule. What's fine in one course can count as cheating in another, so check every time."
Will a detector flag me?	"Detector scores aren't treated as evidence. Follow the policy, and talk to us if you're stuck."
The syllabus doesn't say anything.	"At UMass the default is no AI without permission. I'll ask Professor ___ and we'll post the answer."

When you suspect misuse

Don't accuse, and don't grade on a hunch. Write down what you noticed and why, keep the evidence, and tell the instructor. The Handle It or Hand It to the Instructor? page has the steps.
Never use an AI detector score as evidence. Independent tests found detectors neither accurate nor reliable, and one study found that seven detectors flagged essays by non-native English writers as AI-written more than 60% of the time on average. Many TAs and students in computing are non-native writers.
In office hours, ask students to explain their own code. It supports learning under every policy, and it is the fairest way to see what a student understands.

Code-similarity reports

Many courses run submissions through a similarity tool such as MOSS, JPlag or Gradescope's code similarity report. If you are asked to look at one:

A high score means the code is similar, not that anyone copied. The tools say so themselves: a person has to read the matched code.
Exclude the starter code before running the report, or every student will match the handout.
Expect false positives on small or tightly specified assignments, where many correct solutions look alike. Read the matched regions side by side.
Expect false negatives too. Renaming and reordering can hide copying, and AI-generated code may not match anyone. A low score doesn't prove the work is original.
Never act on the score alone. Write down what matched and where, and send it to the instructor. Don't contact or accuse the student.
Keep reports private. They put students' names side by side.

Why the policy matters for learning

Research on novice programmers suggests AI tools help students who are already doing well, and can leave struggling students with an "illusion of competence": working code they could not have written or explained. Surveys of computing students find debugging is their most common use of AI. Whatever your course's policy, steering students toward explaining and tracing their own code is the help that lasts.

Keep your own view out of it

Students will ask what you think about AI, or tell you the policy is pointless. You can say that reasonable courses differ and that the instructor has chosen this rule for this course. Arguing for or against the policy in section puts students in an awkward spot and makes the next TA's job harder. If you think the policy has a real problem, tell the instructor.