AI may be about to expose one of higher education’s quieter design flaws:
we made too much human interaction optional.
MIT’s new report on AI and education calls for universities to deliberately protect the human side of learning as AI becomes more capable. The reason is easy to see. When students can get an explanation, clarification, second opinion, or even a reasonably convincing answer from GPT in seconds, the incentive to find a study group or walk into (or online) office hours starts to disappear.
I see this from the classroom, not from a committee room. I move from one business school to another and, quite literally, teaching is what I do.
A few years ago, I could spend roughly 50 hours a semester across my classes in voluntary office hours. This past year: zero.
Slack tells me much the same story. I have used it with students for years, filling it with questions, polls, discussions, GIFs, odd and fun things I find online and the occasional late night exchange. Thousands of interactions have gradually dwindled to almost nothing. What remains is mostly administrative: Which file? Which date? Where do we upload this?
But the part that fascinates me most happens downstairs.
I often work in school cafeterias and public spaces between classes, surrounded by student groups doing assignments. I am there to work, but it is impossible not to notice how differently students behave once the classroom structure disappears and no professor is standing over them. That matters, because this is where you see what the system has actually taught them to do on their own. Over time, quite unintentionally, I have become a sort of cafeteria ethnographer, minus the clipboard.
The pattern is increasingly familiar. A group reads the assignment, divides it into pieces, allocates the pieces around the table like a card game, then everyone disappears individually into AI bubbles. Twenty minutes later, they reconvene, almost like waking up from a dream, and stitch the outputs together.
They’re sitting together. But they’re not thinking together. And I think blaming AI gets this backwards.
If a chatbot can remove students from the social life of learning simply because it is faster and easier, then perhaps we never embedded that social life deeply enough in the first place!
Too much teaching is still one way. Courses are short. Syllabi are stuffed. Theory has to be covered. A real debate can swallow half of a three hour class, which makes it wonderfully educational and terribly inconvenient.
So collaboration gets encouraged rather than designed in.
That’s what I have been trying to change.
I lecture far less than I used to. In some sessions, students spend close to 80% of the time working, talking, arguing, testing decisions and getting challenged by me as I move from team to team. Intermediate deliverables create conversations with me. Peer critique forces teams to confront other teams. In entrepreneurship projects, I sometimes make evidence of real conversations with (target) users worth 30% of the grade, and they must show me how those conversations actually changed what they built.
And where the syllabus gives me room, I have been killing off the final PowerPoint presentation too. Students already know how to stand beside slides and read them to people. Instead, I bring in outside practitioners and turn the final presentation into a workshop. Fifteen or twenty minutes with someone who knows the subject, asks awkward questions and does not necessarily accept the first answer. Everything discussed is on the table or the whiteboard, no hiding behind shiny slides that someone else (AI) produced.
Much closer to life, in other words.
Why? Because the workplace students are entering is not an individual prompt box.
It is negotiation, disagreement, persuasion, judgement, collaboration, misunderstanding, compromise and occasionally sitting across from someone who is confidently wrong.
AI can help students prepare for that world. I use it myself, constantly.
But if business schools allow AI to turn learning into increasingly efficient parallel solo work, we risk preparing students to produce excellent answers for organisations where answers were never the hardest part.
The truth is, the report can be generated. But the meeting still has to happen.