LLMs in Learning

The goal of a Boston College education is:

“We help students develop their whole selves by integrating their intellectual, social, ethical, and spiritual growth, so they can discover their passion and pursue lives of purpose and meaning.”

Boston College Formative Education, emphasis added

My goal as your professor is to aid and incentivize you to develop your whole selves within the context of our course, that context being the course's material, your peers, and your professor. I understand this to mean that you should be a changed person by the end of this course, and this should be more than “I now know how to prompt AI to do XYZ.”

If you develop your whole self without using any generative AI, I think this is wonderful. If you use generative AI to help you in this endeavor, I wish to warn you that I have seen it be a polarizing force. I have observed some students flourish as they incorporate AI into their learning. I have observed some students languish badly because they use it as a crutch. The warning I bring is this: I believe generative AI use becomes a crutch by mistake, meaning, most students don’t realize that their use is holding them back.

And how is AI a crutch? Suboptimal usage of AI shortcuts the productive struggle of learning. See this comic below, created by Nick Sousanis of San Francisco State University (source for comic).

The jagged and zigzag pattern is how learning works. To achieve the level of learning expected in a university setting, you must struggle and metaphorically bang your head on a wall first. It's a prerequisite. Asking The Oracle gives you the answer, but It cannot not learn for you.

How can you tell within yourself if your AI usage is turbocharging your learning or hindering you? Here are some clues I invite you to ponder.

Clue 1. Using generative AI to do the assignment for you is a clear violation of Academic Integrity. Work submitted under your name carries the declaration that the work was developed with your own creative forces. Work that does not seem to be your own will be subject to an investigation regarding academic integrity, whether it is generated by AI or whether it comes from a more traditional method of cheating.

Clue 2. Quizzes and examinations rely on you having developed innate skills that you can recall at a moment’s notice. If there is a stark difference between your performance on homework and performance on examinations, it could mean you are relying on generative AI inappropriately. This can also occur if your study skills themselves need improvement. For this, please visit the CFLC and meet with your professor.

Clue 3. Be intentional with how you use generative AI.

Very Bad Prompts

• Solve this problem for me.

• Write solutions to this assignment.

Bad Prompts

• (Before putting in substantial effort) How do I get started on this problem?

• Tell me what to do next in this problem.

Good Prompts

• (After spending considerable effort and time into a problem) Can you give me a hint without giving away the next step?

• Explain how to solve a different but similar problem.

Great Prompts

• Explain this topic/concept for me.

• (After completing an assignment) Generate 3 more similar problems to use for practice.

• (After completing a problem) Upload a photo of your work and ask for critique.

Clue 4. Even if your prompts are extremely responsible, your AI usage might be replacing human interaction. This has been the saddest consequence of the spread of generative AI in my math courses. The standard model for learning difficult mathematical or scientific material used to be meeting up with a regular group of peers at a blackboard and you hash out the assignments together. Everyone is there with a real desire to learn and they are going far deeper than sharing answers. They are trying things out, making mistakes, reading their notes and textbooks... and someone has a eureka moment and then teaches it to their friends right then and there. Once everyone figures out the problem, they all write up their own solution and then move on to the next problem. It's remarkably effective, produces genuine understanding for everyone involved, and it's fun.

Since generative AI became good, I can count the number of times I've seen this occur on one hand. I genuinely believe this is a catastrophic loss to the college experience. You're losing deep and meaningful friendships. Believe me when I say this because I'm a 99th percentile introvert.

Generative AI also replaces interactions with professors. My colleagues universally report a massive reduction (50%-100%) in students visiting their office hours. Your professors want you to come to their scheduled office hours. You will see a different side of your professor in office hours and their efforts can be focused directly on you. This is a huge resource! And, if nothing else, when you need letters of recommendation down the road, who will you ask? I hope it's not a professor who doesn't even remember your face.

I find my reaction to this question to be incredibly insightful for judging any usage of generative AI: 

"What is this replacing?"

Let's conclude this discussion with ChatGPT telling you how to use it, and then I also get ChatGPT to willingly tell on itself a little bit. This was done with ChatGPT 5.6 Sol High. Link to the Chat.