AI for programming assignments: get help, avoid misconduct

First things first: this post is not legal advice. There is no single rule for every university or college. What's allowed differs between institutions, departments, courses and sometimes even between individual assignments. One instructor allows AI for explanations, another bans it completely, a third asks you to declare every bit of help you used.

This is where you find the rules that apply to you:

• The assignment itself and the course page in Canvas, Moodle, Blackboard or a similar system. There is often a section on permitted help or collaboration.
• The course syllabus or module handbook. It tells you how much the assignments count toward your grade, and in some courses whether you need them to sit the final exam. Many instructors also state their AI policy there.
• Your institution's academic integrity policy, and its honour code if it has one. It defines what counts as misconduct, such as plagiarism or unauthorised collaboration, and what can happen as a result.
• Your institution's guidance on generative AI. Many universities and colleges have published their own. Search for it on your institution's website.

If you can't find anything clear, ask the instructor or teaching assistant, ideally in writing by email or in the course forum. A clear answer protects you better than any guess from the class group chat.

Without a specific rule, a simple picture helps. Think of AI help as a scale, from “explains something to me” to “writes it for me”:

1. Getting a topic explained, for example how arrays in Java work.
2. Getting an error message translated into plain language.
3. Getting a hint about where in your own code the problem is.
4. Getting your own code corrected.
5. Having the whole assignment solved and handing it in.

The further down the list, the more likely it becomes a problem. Step 5 is a problem for individual assignments almost everywhere. Where exactly your line is, though, only your course rules can tell you. By the way, the same applies to solutions from classmates or from previous years.

Also check whether an assignment is meant as individual or group work. In a group you can usually think things through together. Taking code from another group is something else. And if the AI only answers a general question without seeing your assignment, you're usually closer to step 1 than to step 5. You can only be sure, though, once you know the rules of your course.

Even without similarity-checking software, copied code is often noticed:

• It doesn't match where the course is. You're on loops, the submission uses streams and generics.
• The style changes. Variable names, indentation and structure suddenly look very different from your earlier submissions.
• Many submissions look alike. When several people paste the same task into the same tool, the solutions often come out similar.
• You can't explain it. In some courses you present your solution in a lab session or answer questions about it.

The most important point is a different one, though: you take the exam on your own. If you let something else do the assignments, you might collect the points and then face questions at the end of term that you've never solved yourself.

These rules work equally well with AI, tutors and study groups. Still check whether your course has stricter requirements.

• Ask about the concept, not the assignment. Instead of pasting the assignment into the chat, ask in general: “How do I return two values from a method?” The basics are in the article on methods and recursion.
• Build your own, smaller example. Understand the idea with it, then apply it to the assignment yourself.
• Write every line yourself. If you see code, close the window and write from memory.
• Declare help when it's required. Some courses want to know which tools you used. Then say so honestly.
• Keep your intermediate versions. Several versions of your file show how your solution came about. That also helps you when you review later.
• Use office hours and lab sessions early. With a concrete question about your own code, you usually get help quickly.

More prompts that give you hints instead of solutions are in learning Java with ChatGPT.

Often you reach for AI because the assignment asks for a big step, and the small steps in between are missing. The lecture shows what a class is. The assignment immediately asks for a program with three classes and a list. The best prevention is to practise the basics beforehand on your own, smaller exercises. Then the assignment becomes an application instead of a puzzle. How to get through weekly assignments without all-nighters is covered in weekly programming assignments.

That's exactly what you can use Jario for. The tasks there aren't your assignments, they're separate exercises, sorted from “Easy” to “Hard”. Before an assignment on classes and objects, for example, you first solve a few small tasks on the same topic. Tests show you immediately whether your code is correct. The AI tutor knows the task, your code and the test result and gives you hints in small steps, but deliberately no solution. Important: the tutor is an AI too. If your course bans AI help on an assignment, that applies to the tutor as well. In that case, use Jario to practise beforehand and solve the assignment on your own.

• The rules are in the assignment, the syllabus, your institution's academic integrity policy and its AI guidance.
• Unclear? Ask your instructor in writing.
• Get things explained, not written. You type the code yourself.
• Practise the basics on your own exercises first, then you'll need less help with the assignment.

How to plan your whole first programming course is shown in how to pass your first programming course.