AI in Education5 min read
Students Are Already Using AI. Universities Need to Catch Up.
The policy debate inside most universities is still about whether to allow AI. Students resolved that question on their own a couple of years ago.
ChatGPT was not built for universities. It was built to be useful to everyone, which means it was built to be optimized for no one in particular — and that design choice has real consequences in academic settings.
Built for Everyone, Optimized for No One
ChatGPT was not built for universities. That's not a criticism — it's just a description of what it is. It was built to be useful to everyone, which means it was optimized for no one in particular. Ask it about macroeconomics and it'll give you a reasonable answer. Ask it about the specific theoretical framework your professor assigned for your intermediate macro course and it'll give you something that sounds reasonable but may have nothing to do with what you actually need to know.
That gap is the core problem with deploying general-purpose AI in higher education. It's not that the tools are bad. It's that they were built for a different context than the one they're being used in.
The Hallucination Problem
The most widely discussed failure mode is hallucination — AI systems confidently stating false information. This happens in all domains, but it's particularly consequential in academic settings where precision matters.
A student asking about a legal case for a law course, a specific historical event for a history seminar, or a particular experimental result for a biology lab is operating in a context where the details matter. Getting the general shape of an answer right but the specifics wrong isn't a minor inconvenience. It's the kind of error that leads to misconceptions that take weeks to undo.
General-purpose AI is trained to produce plausible-sounding text. It has no special obligation to accuracy in any specific domain, and no awareness of what your course actually covered.
The Alignment Problem
Separate from hallucination is a subtler problem: even when general-purpose AI is factually accurate, it may not be teaching the right thing.
Every professor makes choices about how to frame concepts, which theoretical lenses to emphasize, what to include and what to leave out. Those choices are pedagogical decisions — they reflect what the professor believes students need to understand and in what order.
A general-purpose AI knows none of this. It answers questions based on what's broadly true, not based on what a specific course is trying to accomplish. A student in a course that deliberately avoids one framework in favor of developing intuition first gets answers that preempt the scaffolding their professor built. The AI isn't wrong exactly — it's just teaching a different course.
The Governance Gap
The third problem is structural. When students use public AI tools, none of the interaction is visible to the institution. Faculty can't see what their students are asking, can't tell whether AI engagement is helping or hurting understanding, and can't tell when a student's confusion was introduced by an AI interaction two weeks ago.
This is a governance gap in the literal sense: a significant portion of the learning environment is operating without any institutional oversight. Universities have processes for everything — curriculum design, academic support, tutoring, office hours — and then this large unstructured thing where a substantial portion of student-AI interaction happens and nobody has any idea what's occurring.
The Alternative
The answer isn't to pretend general-purpose AI doesn't exist or to try to ban it. The answer is to provide students with something better — AI that knows what their course actually covers, responds in a way their professor actually endorses, and generates visibility into how it's being used rather than operating in a black box.
General-purpose AI isn't going away. The question is whether universities will remain dependent on it, or build something designed for the context they're actually operating in.
Share