Project R-16716

Customized LLM-environment to support learning processes in bachelor's programs. (Education)

Students increasingly use AI tools — and Large Language Models in particular — during their learning processes, yet the current setup gives rise to a number of issues. A structural inequality arises between students who can afford (costly) paid subscriptions, and the rest, while the tools that are used typically take on an oracle role: they provide an answer immediately, without further guidance. In this IOP we propose a custom service that addresses these concerns in a sustainable manner, in much the same way as other campus-wide facilities such as internet or e-mail. Concretely, this will be a uniform interface through which students can access a broad set of AI models and compare their output — contrary to the current situation in which every student has a personal account on a single LLM. By building on existing open-source initiatives and tailoring them to the educational context, the cost model shifts from flat-fee subscriptions to a pay-per-request model. The tools will be adapted to take on a coaching role — a virtual tutor rather than an oracle — while the teaching team, with explicit attention for privacy, gains insight into the prompt behaviour of students through anonymised analytics. In this way we increase the students' AI literacy, and obtain a better estimate of the actual cost of providing such facilities. The service is initially rolled out in the bachelor of Informatics, and in a second stage also in Biology, so as to evaluate its portability to another context.

01 September 2026 - 30 September 2027