Every Syllabus for Itself.

All of my classes this semester are creative writing courses and thus all of them on their respective syllabi, list AI use under academic dishonesty, flat, without qualification. STEM courses—generally speaking—are more open to the use of AI, especially courses that I’ve taken in Computer Science. Some other courses though just use the boilerplate DePauw language on AI use. Three and a half years into a technology that has rearranged how students across American universities study, write, research, and sometimes cheat, the professors teaching me this spring have arrived at three different answers. And they all think they are right.

According to a March 2026 research report commissioned by DePauw through the Lilly Endowment's Artificial Intelligence in Higher Education planning grant, 25 percent of DePauw students who have used AI since fall 2025 have used it in only one course; 30 percent in two or three; 19 percent in four or more. The numbers describe what the syllabi add up to. DePauw does have institutional AI guidelines, updated as recently as April 7. They address data privacy, content responsibility, and security risks. On AI in coursework, however, they defer to faculty, who are asked to set their own policies. The institutional stance on AI in the classroom is that each faculty member should have one.

There are reasons for that. According to the same report, only 19 percent of DePauw students, faculty, and staff believe AI provides accurate, trustworthy information, compared with 65 percent of parents of prospective students and seventy-five percent of prospective students themselves. Ninety-three percent of DePauw faculty express concern about ethical issues in AI systems. Faculty hesitancy is not a failure to keep up. It is the opposite. Faculty have looked at these tools up close and concluded that they are unreliable, biased, and corrosive to the kind of intellectual work a liberal arts education is supposed to produce. When a professor bans ChatGPT from their classroom, they are defending something—the slow, frustrating, voice-shaping work of writing a paragraph from scratch—that they know is at stake.

That defense deserves to be taken seriously, and in its essentials I believe it is correct. The struggle of writing badly before writing well is part of how any writer becomes one, and a tool that smooths out that struggle in exchange for a competent-looking paragraph is not neutral. But the struggle is not what is being defended on the ground. What is being defended on the ground is whichever policy each individual professor arrives at, enforces in whichever way that professor chooses, with no guarantee that the professor across the hall is defending anything similar. Last semester, I took a creative writing course where the professor got rid of papers outright and instead had us write in blue books. In a computer science course, that same semester, AI usage was allowed and the only thing I had to do was cite it and show what content was made using AI. Neither of these approaches are wrong, but it’s a sign of a divergence in the humanities and the sciences. The same report found that 55 percent of prospective undergraduate students—the class that is the future of DePauw—use AI tools several times a day. They arrive already fluent in something their professors haven't yet agreed on how to name. The cost of that disagreement is not equal. The report found that only 39 percent of first-generation prospective students use AI several times daily, compared with 67 percent of their non-first-generation peers. Eighty-one percent of first-generation students feel confident writing effective prompts, compared with 93 percent of non-first-generation students. The students most likely to misread a syllabus are the ones who already arrive with less fluency in these tools, less confidence in their own reading of institutional rules, and more to lose if they guess wrong. When the institution declines to set a coherent policy, the ambiguity costs something that specific groups absorb.

Some reading this might think of my use of the em dash (—) as a sure sign of AI use—something that irks me to this day, and is something that shows how pervasive AI paranoia has become. The em dash has existed long before artificial intelligence, long before any of us living today, but it’s become infamous for being a big part of AI-produced content. This leads me into my next point: It should be every higher-education institution’s responsibility to teach their students how to use these tools, for what purposes, for when, and how to spot its use in the wild. We’re past the point of using tools like GPTZero—which present as being infallible AI “detectors,” but in reality are nothing more than tools that guess where the content came from. These tools claim that the United States Constitution was written by an AI agent—I think that proves my point for me. The way one can actually detect AI use is by reading AI content. Wikipedia released an article a few weeks ago which outlines observations from experts on what AI systems rely on when producing content—we should be teaching this to students. This is obviously also a double-edged sword—teaching students how to detect AI use, will surely make them better at masking their own AI use for things like cheating.

I am not asking DePauw to mandate AI use, or to ban it, or to resolve a technological question that no university has yet resolved. I am asking for the baseline a student should expect from a tuition-funded institution: that the rules of permissible work are shared across a department, published in a way a first-year student can find, and consistent enough that a student taking four courses in a semester is not being governed by four different and sometimes contradictory regimes. It’s also telling when the aforementioned DePauw institutional AI guidelines were produced with the help of AI tools too—”Portions of this document were produced with the assistance of Claude.AI”. The DePauw report itself acknowledges that the absence of clear institutional policy produces confusion, and that students across campus are actively asking for shared guidance. The syllabi on my desk will be there through the end of the semester, and the contradictions in them will stay contradictions. I will graduate in May having learned, among other things, to read each professor's AI policy like a legal document and to adjust my conduct accordingly. That is a kind of education, too. It is just not the one the catalog advertises.