
Every college in America now faces the same question: what is a university for when a chatbot can write the essay, solve the problem set, and summarize the reading in seconds?
Most institutions have answered by bolting AI onto the old model. They add AI tutors, AI graders, and AI "teaching assistants," and they keep the lecture halls, the semesters, and the price tag.
We think that gets the question backwards. AI changes what a degree has to prove. A diploma used to signal that a student could produce certain work. Now a machine can produce that work. So a degree has to signal something AI can't fake: that a human being actually learned, can think independently, and can do real work in the world.
That's our answer to the debate over AI in higher education, and it's what we mean by a post-AI university. It doesn't ignore AI. It assumes AI is everywhere, and it builds education around the part of learning that only happens inside a human mind.
Is AI Making Us Dumber? Learning Takes Effort, and AI Removes It
Learning scientists have known for decades that effort drives learning. Robert and Elizabeth Bjork call these productive struggles "desirable difficulties": retrieving an idea from memory, working a problem before seeing the answer, and explaining a concept in your own words all feel harder in the moment and produce far more durable learning.
Generative AI removes exactly that struggle. Researchers call it cognitive offloading, and new research shows the cost.
- Students score better with AI, then worse without it. In a randomized trial with nearly 1,000 high school math students, researchers at the University of Pennsylvania gave some students access to a GPT-4 tutor during practice. Those students solved 48% more practice problems correctly. But when the AI went away, students who had used the standard chatbot scored 17% lower on the exam than students who never had it.
- Heavy AI use tracks with weaker critical thinking. A 2025 study of 666 participants found a significant negative correlation between frequent AI tool use and critical thinking scores, with cognitive offloading as the link. Younger participants showed the highest dependence and the lowest scores.
- Knowledge workers report thinking less. A Microsoft Research and Carnegie Mellon survey of 319 knowledge workers found that the more people trusted generative AI, the less critical thinking they reported applying to their work.
- The brain shows it, too. An MIT Media Lab study recorded EEG activity while participants wrote essays. Participants who wrote with ChatGPT showed the weakest brain connectivity of any group, and many struggled to quote from essays they had just "written."
(We dig deeper into this research in Is AI Making Us Dumber? What the Research Says.)
None of this means AI is bad. It means AI is a powerful tool for people who already know how to think, and a dangerous shortcut for people who are still learning how.
What a Post-AI University Does Differently
A post-AI university starts from a simple principle: No AI teaches at Elton. Every tutorial, every piece of feedback, and every grade comes from a human expert who knows the student and their work.
We build on three practices that AI can't hollow out.
1-on-1 human teaching. In 1984, Benjamin Bloom found that students tutored one-on-one performed two standard deviations better than students in a conventional classroom, his famous "2 sigma problem." More recent evidence confirms the core finding: a 2020 meta-analysis of 96 randomized trials found that tutoring raised learning by 0.37 standard deviations on average, one of the largest effects in education research. That study covered PreK–12 tutoring, and we build the same principle into university-level learning. Every Elton student works directly with expert professors and professional coaches.
Conversation as assessment. A chatbot can draft a paper. It can't sit across from a professor and defend an argument, answer follow-up questions, and revise its thinking in real time. When students explain, defend, and extend their ideas in live conversation with an expert, the learning shows, and so does its absence.
Real-world work. Elton students build projects, portfolios, and professional experience alongside their studies. Work that lives in the world, with real users, clients, and stakes, demonstrates capability that no transcript of AI-assisted assignments can.
A Post-AI University Still Teaches AI
Post-AI doesn't mean anti-AI. Our graduates will work with AI every day, and they need to understand how it works, where it fails, and when to trust it. Students can study AI engineering and AI-focused business at Elton, and every student learns to evaluate machine-generated claims with the same rigor they apply to human ones.
The difference is the order of operations. First, students build the knowledge and judgment that let them think for themselves. Then they use AI to amplify that thinking, not replace it.
The Future of Higher Education Is Human
A democracy needs citizens who can weigh evidence, spot a bad argument, and change their minds for good reasons. An economy needs workers who can solve problems no one has solved before. Communities need people who can listen, reason together, and act.
If a generation outsources its thinking during the years it's supposed to learn how to think, we all lose those capacities. A post-AI university protects them. It treats human attention, curiosity, and judgment as the scarce and precious resources they've become.
Learn With Humans at Elton
Elton University is a nonprofit, evidence-based university where every student learns 1-on-1 with expert human professors, online from anywhere or in person in San Francisco, on a schedule built around their life. Our undergraduate, master's, doctoral, and certificate programs are enrolling now.
Explore our academics, read our FAQ, or book a free info session to see whether a post-AI education is right for you.
Sources
- Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26).
- Bjork, E. L., & Bjork, R. A. (2011). Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. In Psychology and the Real World.
- Bloom, B. S. (1984). The 2 sigma problem: The search for methods of group instruction as effective as one-to-one tutoring. Educational Researcher, 13(6), 4–16.
- Gerlich, M. (2025). AI tools in society: Impacts on cognitive offloading and the future of critical thinking. Societies, 15(1), 6.
- Kosmyna, N., et al. (2025). Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task. MIT Media Lab, arXiv preprint.
- Lee, H.-P., et al. (2025). The impact of generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. CHI 2025.
- Nickow, A., Oreopoulos, P., & Quan, V. (2020). The impressive effects of tutoring on PreK-12 learning: A systematic review and meta-analysis of the experimental evidence. NBER Working Paper 27476.