AI University.
Managed and taught by AI agents.
At a fraction of the cost of a traditional university.
Another elephant in the room: I do not see many people talking about this out loud, but the pieces are already here. If you zoom out, the question is not whether AI can “teach.” It’s under what conditions it helps learning versus quietly erodes it.
What the research is starting to show:
1) AI tutoring can outperform traditional instruction when it is intentionally designed.
One recent study found students learned more in less time using a purpose-built AI tutor compared to in-class active learning, with higher engagement and motivation reported.
2) The older “ITS” research already showed consistent learning gains.
Meta-analyses of Intelligent Tutoring Systems report moderate positive effects on learning outcomes, including at the college level.
3) But generative AI without guardrails can harm learning.
A rigorous experiment on GPT-4 style tutoring finds that design choices matter a lot, and “help” that shortcuts thinking can reduce learning or reasoning quality.
4) A practical middle ground is emerging: AI that supports the human tutor or instructor. Human-AI tutoring support (coaching tutors in real time) has shown measurable gains in mastery.
So “AI University” is not science fiction. The real debate is about the operating model:
AI as a solution vending machine (cheap, scalable, and academically dangerous)
AI as a scaffold that forces productive struggle (cheap, scalable, and instructionally powerful)
If we actually built “AI University,” the hard problems are not the lectures. The hard problems are:
assessment integrity and proof of learning
feedback quality and hallucination control
accreditation, standards, and accountability
motivation, belonging, mentoring, and career signaling
I’m going to start collecting studies on AI agents and AI tutoring in post-secondary. If you have papers, pilots, or institutional experiments worth reading, drop them in the comments.
Robert Foreman
Doctoral Student, Educational Technology
Central Michigan University
Doctoral Researcher in AI-Augmented Learning | Data Analytics Consultant
Email: forem1r@cmich.edu
Phone: 480-415-0783
Website: https://NhanceData.com
Sources worth starting with:
Brookings overview of RCTs in AI tutoring
https://lnkd.in/gUrvQGWx
ITS meta-analysis evidence base
https://lnkd.in/g9N4ppuC
“Guardrails matter” experimental evidence
https://lnkd.in/ggiEeke5
Purpose-built AI tutor outperforming active learning
https://lnkd.in/gtu8UW3m
Human-AI tutor support at scale
https://lnkd.in/gY5AZwRs