Teaching with AI: UvA AI Chat Sparks Independent Learning

After telling her students that they were allowed to use the UvA AI chat during the practical sessions of this course, they responded surprisingly calm. Christa laughs as she recalls thinking she was announcing something innovative, while the students explained that many of them already used AI on a daily basis, so it was hardly news for them.

To prevent students from relying blindly on AI, Christa devised several strategies to increase the likelihood that students would use the chat responsibly while still being challenged:

• Christa deliberately explored the limits of AI. By investigating what AI was not yet good at, she designed assignments around those weaknesses. For example, AI was previously not very good at analysing images. Sometimes she asked students to sketch something on paper and take a photo of it.
• During the exam, AI was not allowed. “Students can passively paste assignments into AI and see what comes out, but they will inevitably run into problems during the exam, where AI is not permitted,” says Christa. “The same applies to presentations: they may use AI to brainstorm, but they must be able to present and defend their work themselves.”
• A major advantage of AI assistance during programming was the built-in reality check. If AI generates code containing an error, you notice it immediately when you run it: the computer returns an error message. That is typical of programming. When writing an essay, this is more difficult, because errors in logic or argumentation do not trigger automatic warnings. These error messages taught students to engage critically with AI output. “As a researcher, asking the right questions is crucial. You only get a valuable answer if you ask a good question.” This is an essential skill in science.
• From the very first lecture, Christa facilitated open discussions about where AI could help and how students could remain critical. “There was quite a bit of variation among students. Some were rather sceptical and cautious about AI, while others used AI almost all day long.” Within the course, students were free to choose whether or not they wanted to use AI.

“Not every lecturer will agree with me,” she says, “but in this course it’s not necessarily about students being able to write flawless programming code independently. It’s about knowledge, understanding analyses, and finding solutions.” She also emphasises that AI is here and now. “AI is already present all around us. I think many employers expect graduates to be able to work well with AI, and some even see it as an advantage.”

Many colleagues find AI intimidating: what if students let the UvA AI chat do all the work? Christa recognised these doubts but was glad she had conducted the AI pilot. Despite the controversy and uncertainty surrounding the topic, she saw clear benefits emerging from the results.

The pilot showed, for example, that students were able to work through the material more independently and efficiently. Moreover, they could complete more complex assignments involving larger datasets. Students themselves also reported positive experiences with using the chat, as it helped them better understand the course material.

In addition, Christa noticed that the role of the lecturer did not disappear. “What also surprised me was that they still needed me,” she says. “ ‘AI says this, what do you think?’ they would ask. They also still came to us for a pep talk or a joke in between.”

It did prove difficult to determine whether students had actually become better at coding, but thus was not the goal of the course as emphasised by Christa: “The focus was really on understanding the material. AI requires lecturers to rethink assignments and guide students differently.” She adds: “I remember when I was younger and the spell checker in Word was introduced. At first, it was seen as cheating. Nowadays, in a professional context, you can hardly afford to submit a text that hasn’t been spell-checked. I can imagine AI following a similar path: at first we may see it as cheating or a threat to knowledge development, but over time it may become normal and we may assess skills in different ways.”

Christa is a sleep researcher who conducts sleep measurements at night and teaches during the day. A busy schedule that leaves little room to explore educational ideas. Thanks to the grant, time was freed up in her teaching schedule, allowing her to put ideas she had been considering for some time into practice. Receiving the grant felt like a unique opportunity. “I truly felt motivated to make something special out of this pilot. I found it incredibly exciting to work on and it was a great incentive to follow through.”

She can easily imagine that many lecturers, like herself, have ideas to improve their teaching and try new approaches, but simply lack the time to start, or to keep going. “If you have an idea, go for it. It’s really worth it and very rewarding.”