In STEM, the temptation is sharper: the problem has a right answer, and the AI will give it to you. The question is whether you can reproduce that answer on an exam, or in a lab, or in a job — without the AI in the loop.
The math-and-science exception
AI is useful in STEM for explaining concepts, walking through similar example problems, checking your logic after you have attempted a problem, and helping you debug code you have written. AI is dangerous when it does the problem for you — because the exam will not have AI in it, and the skill you were supposed to build doesn't build itself.
The rule for problem sets
Attempt the problem yourself first. Fully. Even if you get it wrong. Then — and only then — ask the AI to either (a) explain the concept you got stuck on, or (b) check your work and point out where your reasoning broke down. Never paste a problem and copy down the answer. That study session taught you nothing.
STEM Scenario · 20 XP
You're stuck on a differential equations problem set. Problem 3 has you baffled. What's the right sequence?
The second option is where learning happens. Getting stuck, naming the step you can't do, getting a different example, and then finishing your own problem — that sequence builds a skill you will still have next week. Copying the answer builds nothing and sets up a worse problem on the exam.
Coding courses
Students in intro CS courses face the most acute version of the AI-assisted-vs-replaced problem. AI will cheerfully write your entire assignment in the exact language and style your course requires. The exam — and every technical interview you ever sit for — will require you to write code yourself, from scratch, without AI completing your sentences.
The workflow that builds actual coding skill: write your own attempt first. When it doesn't work, ask the AI to explain the concept underlying the bug — not to fix the bug. Then go fix it yourself. The next time you hit the same bug, you will recognize it.
Prompt · 15 XP
The Coding Concept Prompt
I am working on a [language] assignment that requires me to [task]. Here is what I understand about the concept: [describe your understanding]. I am stuck on [specific issue].
Please explain the concept behind what's going wrong, using a small, different example (not my code). I will write my own solution from your explanation.
Lab reports and technical writing
A lab report is an argument: here is what we tried to find out, here is what we did, here is what we found, here is what it means. Generating the argument from AI defeats the purpose of the lab. Using AI to pressure-test your data interpretation, check your statistical reasoning, or critique your draft — that makes your report stronger.
Check · 15 XP
Which use of AI in a lab report would violate most academic integrity policies?
The interpretation of data is the intellectual work of the course. When the AI interprets the data and writes the report, you have not learned the method, not learned the field, and submitted work that is not yours. The other three options use AI to improve something you did. That's the line.