Rethinking AI in Academia
How UvA AI Chat Aims to Facilitate Learning, Not Replace It
Brief Introduction To AISO’s New Blog Section
We’re excited to introduce a brand new addition to AISO - our very own blog series!
I’m Karina, and I’ll be the writer behind this and future posts, all created specifically for you, the AISO community. This space is here to help you explore AI concepts, tools, trends, and student perspectives - whether you're just getting started or already building projects. Have ideas, feedback, or questions? Feel free to reach out at karinakalicka@gmail.com. Hope you enjoy the very first post of our new blog series, stay tuned for much more to come!
A Changing Landscape for Learning
We’ve been seeing this trend for around some time now. It’s time to take a look at the solutions.
As artificial intelligence becomes increasingly embedded in education, universities around the world are facing a complex challenge: how to integrate AI tools in a way that enhances learning rather than replaces it.
The question universities now face isn’t whether to use AI, but how.
The University of Amsterdam (UvA) has decided to take that question seriously. Through its Teaching and Learning Centre (TLC), it launched UvA AI Chat: a conversational AI tool designed not to outsmart students and replace their thinking, but to help them learn better.
It’s not meant to compete with ChatGPT or other commercial giants. In fact, it exists because of them - as a deliberate counterweight to tools that make it all too easy to outsource thinking.
“We wanted to make sure the tools we offer don’t take away the opportunity to learn, but actually facilitate it,” says the TLC development team.
That one sentence captures the essence of what UvA is trying to do: protect learning from the temptation of automation
When Machines Start Doing the Homework
Every university is seeing the same pattern. AI has become a study companion, a writing partner, and sometimes, a ghost author. Essays, research summaries, even lab reports - all can be generated in minutes. The speed is intoxicating.
But something is quietly lost in that process. When students start to use AI not to understand but to submit, the human act of learning begins to erode.
That’s the paradox at the center of AI in education: the same tools that can deepen understanding can also replace it.
UvA AI Chat was built around that dilemma. For example, its Study Mode is a subtle but crucial feature, it encourages engagement, not shortcuts. The goal is not to provide answers, but to build understanding.
Privacy Before Power
Behind every AI system is a moral decision: who owns the data?
In most cases, the answer isn’t the user. Commercial AI tools can retain conversations, learn from them, and feed them into massive datasets. That’s efficient for innovation, but problematic for education, where intellectual property and academic integrity still matter. A lot.
UvA AI Chat takes the opposite approach. It’s privacy-first. All inputs are anonymized before being sent to any model. Nothing is stored, nothing is used for retraining, and professors can’t access what students write.
Another quiet innovation of UvA AI Chat is that the system is model-agnostic. Instead of binding the university to one corporate AI provider, UvA lets students choose between several models depending on their task. That flexibility is rare, and it is intentionally academic.
Still, building something this independent is hard.
“We only have a handful of student programmers and technical staff. Hosting and maintaining a self-contained AI infrastructure at university scale is a major challenge,” one team member admitted.
Students in the Loop
If AI is to shape education, it should do so with students, not for them. That idea became the basis for the AI Impact Ambassador Programme, created in collaboration with the AI Student Organisation (AISO).
The partnership between TLC and AISO was not accidental. TLC had been searching for a way to connect more directly with students, while AISO, already a hub for student-led AI discussion and experimentation, wanted to make sure that young voices helped guide institutional decisions. Together, they designed the Ambassador Programme as a bridge: TLC would provide access, mentorship, and early prototypes, while AISO would bring in the honest, unfiltered feedback that only students can give.
The initiative gives students early access to new versions of UvA AI Chat and asks for their honest feedback. Not just about usability, but about ethics, learning behavior, and relevance.
So far, some pilot ideas have shaped some interesting experiments:
• A memory feature that personalizes answers over time.
• Mind-map generation for visual thinkers.
• Side-by-side comparisons for studying contrasting perspectives.
“We needed honest, open feedback,” says the TLC team. “In return, students get to see their input directly shape what we build next.”
The ambassadors aren’t beta testers. They’re co-authors in the design of what could become a new kind of academic infrastructure.
A Prototype for the Netherlands… and Beyond
At the moment, UvA AI Chat serves both the University of Amsterdam (UvA) and the Amsterdam University of Applied Sciences (HvA). But the idea is already moving beyond city limits.
Discussions are underway with other Dutch universities, and a few in Belgium, about adopting similar systems or sharing the underlying framework. If that collaboration takes root, it could evolve into something much larger: a national network of academic AI tools built around privacy, transparency, and pedagogy rather than profit.
Such a movement would also reinforce a distinctly European stance on AI: one that values control, clarity, and the preservation of human learning over convenience.
The Friction of Change
For all its idealism, UvA AI Chat faces a familiar problem: people often resist change. Students often compare it to ChatGPT, find it less “fluent,” and move on. But that fluency, the developers argue, is precisely the problem. It trains users to skip thinking.
“There’s a small step everyone needs to take,” the team said. “The system improves with consistent use - but that requires patience.”
Patience, however, is in short supply in a world of instant answers. And yet, the effort to make AI tools that teach rather than tell may be one of the most important educational experiments of our time.
Rethinking What Counts as Learning
The future of AI in education isn’t about banning or embracing it. It’s about redefining what counts as learning in the first place.
UvA’s experiment shows that universities can design technology that supports curiosity and reflection instead of mere efficiency. It also shows how difficult that is; ethically, technically, and culturally.
But perhaps difficulty is the point.
TLC sees that together with students, we can define when AI is helpful, and when it becomes harmful. That awareness is the foundation for a responsible academic future.
If that awareness spreads - across Amsterdam, the Netherlands, and beyond - it could spark a shift in how universities everywhere approach AI: not as a replacement for learning, not as a way to delegate thinking, but as a mirror that helps us see how we learn.