The night I let AI handle my midterm essay outline, I actually learned more

At 1:47 a.m., on the desk in my Logan Square apartment sat a cold cup of coffee and an open philosophy textbook. The midterm essay was due in two days, and the topic was "The Ethical Status of AI in Artistic Creation." I was reading the relevant chapter for the third time, and the more I read, the more tangled my thoughts got. That's when the impulse hit. What if I just asked an LLM to draft an outline?

In my defense, I've always been the impulsive type. Last semester, I bought a used bike for $85 on a whim in the middle of the night and regretted it the next day. This felt similar. "Just do it now, think about it later." So I opened a chat window and typed in a prompt.

"Write me an outline for a philosophy essay on whether AI-generated art can be recognized as genuine creation. Include arguments for and against."

What the model produced was clean. So clean, in fact, that it felt off.

The problem was that it was too smooth

The outline was divided into an introduction, three body sections, and a conclusion, with each item accompanied by sentences like "Some scholars argue that…" But as I read through it, something bugged me. The arguments for and against mirrored each other with perfect symmetry. Next to "AI is a tool and therefore not a creative agent" sat "AI is a new form of creative agent." Both sides sounded plausible, but both only scratched the surface.

I immediately started pushing back. "Wait, isn't the very definition of 'tool' the problem here? If you assume that a hammer and a large language model belong to the same category of 'tool,' doesn't the argument fall apart from the start?" I kept going like this in the chat window, challenging one point after another.

The model instantly agreed with my objections and spit out a revised outline. But the revision was equally shallow. This time, it simply added an item saying "The definition of 'tool' must be established first," with no depth as to why that definition matters. So I pushed back again. It agreed again, and again it only patched the surface.

The LLM as a sparring partner

After about an hour of this back-and-forth, I realized something. I had long since dropped any intention of using the model's outline. Instead, I'd been picking apart its surface-level arguments one by one and, in the process, organizing my own position. The parts I had glossed over with a simple "sure, okay" while reading the textbook suddenly demanded precise explanation when I tried to tell the model "That's not right." My thinking got sharper through that process.

Here's an example. The model claimed, "AI art is indirect creation because it reflects the intentions of the human artist." I was about to agree, but then — wait, does that mean a photo taken on a camera's auto mode is also "indirect creation"? Does the mere act of pressing the shutter make you a creator? When I posed this question to the model, it immediately agreed again. Way too easily. A real debate opponent would have fired back, "No, that's a different issue entirely."

And that's the funny part — the model's limitations were actually educational. Because it agreed so readily, I had to construct the counterarguments myself. I ended up playing both sides on my own.

So what happened to the essay

Around 3 a.m., I scrapped the model's outline entirely and wrote my own from scratch. The core arguments came from the textbook, but the structure and counterarguments were things I'd refined through my battle with the model. After I submitted it, one of the professor's comments read, "The tension between the arguments is well sustained" — and honestly, that tension was the product of repeatedly telling an AI "That's not right" in the middle of the night.

If you ask me whether I'd recommend LLMs as a study tool — I wouldn't call it a recommendation exactly. What I stumbled into that night was this: the smoother and more plausible the answer, the more it trains you to ask, "Is that really true?" And that muscle for questioning is exactly the skill a philosophy essay actually demands.

If I end up doing something impulsive again, I can only hope it works out this well. That $85 bike is still collecting dust in the hallway.