Introducing UvA AI chat for teachers
Full event registration, chapter markers are available on the video timeline.
Last July, Prof. Jessica Taylor Piotrowski and Dr. Sharon Klinkenberg, co-directors of the FMG Teaching & Learning Centre. introduced the new UvA AIchat – which will be available for student use as of September 1 (teachers already have it). Here you will find a summary of the information presented as well as the slide deck, and Q&A.
With the launch of UvA AIchat this semester, we want to prepare teachers for what’s ahead. It’s important to emphasize that if you are teaching within FMG and prefer not to use AI in your courses, there is absolutely no expectation for you to do so.
What Does the UvA AIchat Do?
Short demo on how to use UvA AIchat.
Teachers can already try it out via aichat.uva.nl. The tool allows you to chat with an AI assistant and ask any kind of question — from something practical like “How can I make my teaching less stressful?” to more content-specific prompts. You can choose between different models, upload documents, or point the chatbot to a website for reference. Importantly, all data stays private and secure, and nothing you share will be used to train future AI models. To make things easier, the chatbot itself includes a getting started guide, an e-learning module, and feedback forms you can use at any time
Of course, Sharon couldn’t resist geeking out about how large language models actually work (you can check out his full explanation in the video — we’ll spare you the deep dive here). What’s important to know is this: UvA AIchat uses pre-trained models that are fixed. That means it does not learn from what you type in, and your prompts will never become part of the model’s knowledge. But the models aren’t perfect. Because they’re trained on existing data, some of the biases in that data can occasionally show up in the answers they generate.
If you’d like to bring AI into your teaching, our advice is simple: for now, explore it, but don’t rely on it. Think of UvA AIchat as a tool to spark ideas and support your work — not as a replacement for your expertise.
Planning to try it out in your courses next semester? You can book a session with our AI experts , who are happy to help you brainstorm, prepare, and shape your ideas into practice.
How You Might Use the UvA AIchat
One of the easiest ways to start is to treat AIchat as a sparring partner. For example, you might prompt it with “Help me set up a lesson plan for my tutorial on topic X. I have 2 hours available and I want at least two activating class activities. Prompts like this can quickly jumpstart your ideas. From our experiments, we’ve also seen AIchat work well as a writing coach for students (helping them refine drafts) and a study tool (generating key questions students should be able to answer based on course content).
When introducing it to students, it’s important to stress that the goal of using AI is to support the learning process — not just to get to the “right” answer.
In the assessment space, AIchat can generate a starting point for multiple-choice questions from a specific book chapter; suggest gentler ways of phrasing feedback on papers; or inspire ideas for rubrics.
Practical use cases
And, if you do not allow AI in writing assignments, keep in mind that thoughtful use of AI is not detectable. One useful strategy is to pressure-test your assignments in UvA AIchat. Simply paste your assignment into the chatbot and see what it comes up with. Requiring students to reference course-specific materials or lecture content will at least make it harder for AI to produce a strong answer.
📌 Lastly, don’t forget – AI cannot be used to grade student work and you should always keep a close line with your friendly teaching directors and examination board peers before bringing AI to the party.
Things to Keep in Mind
- Guardrails can be bumpy. We have noticed that the system’s built-in restraints can sometimes be surprising. Of course, it makes sense that guardrails are in place to block harmful content (e.g. instructions for self-harm). But those same guardrails may occasionally interfere with questions that are perfectly valid in an educational context. This is an ongoing conversation within our community about our educational values, so please share your feedback (in the AI tool there’s a feedback option) if you encounter inconsistencies.
- Sustainability matters. No need to thank the chatbot after every session 😉. While most of AI’s environmental impact comes from the training phase, it’s still good to use the tool purposefully. For example, asking a large model to do a simple calculation is massive environmental overkill. That said, we don’t yet have clear insights into how different models compare in terms of sustainability — or how best to communicate that impact to users. This is still to come, but in the meanwhile, it’s a great dialogue to have with your students.
- Resources and support. There are already plenty of support pages to help you get started. In addition to the UvA AI policy document and the general AI page for employees, we’ve listed more resources on the right.
Resourses
- e-learning: Getting started with UvA AI Chat
- Manual: UvA AI Chat
- Website: Teaching & GenAI
- Workshop: GenAI in Assessment
- Workshop: Introduction to AI for teacher
- e-learning: Impact of GenAI on higher education (for lecturers)
- Track: BKO+ Track – Responsible AI in Education
e-learning (for students): Responsible use of GenAI in higher education
Let’s Figure This Out Together
AI in education is still very new, and we are all figuring it out together. Whether you’re curious to experiment or prefer to wait and see, please know that there’s no one right way forward. At the TLC, we’re here to support you, answer questions, and learn alongside you. Let’s explore how UvA AIchat can best serve our teaching and our students — together.
Session with Piotrowski & Klinkenberg – Q&A
Q: With regard to the level playing field for students, are there edge cases where students who pay for ChatGPT benefit more compared to UvA AIchat?
ChatGPT sometimes lets you dive deeper into the reasoning behind certain choices — for example, when discussing racial bias in examples. UvA AIchat may simply stop responding at a certain point. If you’re working with edge cases, you’ll need to experiment a bit yourself to make sure it aligns with your educational goals.
Q: Should we teach students not to use UvA AIchat for fun, just to be mindful of the environmental impact?
We have an e-learning module available for students on the responsible use of AI. You can also encourage them to offset their AI use with sustainable habits, such as avoiding standby mode on energy-draining devices or choosing train travel over flying for holidays. In general, it’s an excellent conversation to have in the classroom as students consider their mindful use of AI.
Q: How does the UvA AI chatbot compare to commercial products in terms of preparing students for the labor market?
When it comes to reflecting critically on what AI can and cannot do, UvA AIchat is quite similar. However, students need a solid knowledge base to be able to reflect meaningfully on its responses.
Q: Are we able to fine-tune the models?
From the development team: if you know how to do this yourself, we can set you up. We don’t currently have the staff resources to fine-tune models ourselves.
Q: From a value-based perspective, will the University continue to have a discussion about the role of AI in our lives and whether it might attempt to ‘replace us’?
Yes, these discussions will take place. The EU AI Act requires teachers to be AI-competent, but what that means in practice is still under debate. A taskforce is working on this at UvA, but there won’t be a simple or fast answer. If you have thoughts on this, reach out so we can begin this shared dialogue as a community.
Q: Shouldn’t there also be a program for students on why they come to study at UvA?
We believe that incorporating self-regulated learning into our teaching will help students use AI responsibly and also reflect on their own educational journey. There will also be new courses on generative AI, with one planned by IIS to launch this February taught be teachers in the FMG.
Q: Is there going to be some sort of API access?
From the development team: there already is. Just reach out to us and we’ll help you get set up.