MiNI team at the prestigious IJCAI-ECAI conference

The solution developed by the MiNI students was accepted for the Demonstrations Track, placing it among just 50 papers selected from 186 submissions from around the world. The paper is entitled ‘From Surface Learning to Deep Understanding: A Grounded AI Tutoring System for Moodle’. The team created a modular plug-in for Moodle, the popular learning platform, using retrieval-augmented generation (RAG). The project aims to counteract surface learning and encourage a deep understanding of concepts through the Socratic method. The publication was authored by Anna Ostrowska, Michał Kukla, Gabriela Majstrak, Jan Opala, Sebastian Pergała, Jan Skwarek and Anna Wróblewska, PhD.

"Put simply, we created a system embedded in the Moodle/LeOn e-learning platform that works in two ways and supports both sides of the educational process. It can be used in any school or university that wants to support teaching with modern technology and automate teachers’ time-consuming tasks while retaining full control over the reliability of the materials," explains Michał Kukla, a student at the WUT Faculty of Mathematics and Information Science.

MiNI team at the prestigious IJCAI-ECAI conference

For school and university students, the solution proposed by our team serves as an intelligent study assistant. It provides an interactive chatbot that can operate in three modes: Quick Mode for fast facts, Deep Understanding based on the Socratic approach, and Exam Coach.

"Instead of searching through dozens of presentation slides for a single topic, a student can ask the chatbot a question. Crucially, the system does not simply provide ready-made answers. In one mode, it uses the Socratic method: it guides the student towards the answer and encourages them to think. At the same time, it relies exclusively on verified materials supplied by the lecturer - books and lecture notes - so we can be confident that the AI neither makes things up nor hallucinates."

For school and university teachers, it is an automation tool that offers substantial time savings. It automates the preparation of tests and materials, but uses a human-in-the-loop architecture, meaning that generated content must be approved by a person before publication.

"Using uploaded files, such as a PDF of presentation slides, the system can automatically generate tests or supplementary materials. Crucially, however, the AI is not left to operate entirely on its own. The lecturer has a dedicated panel where they can review and edit the generated materials before students see them. We call this a human-in-the-loop approach. Of course, every generated quiz or resource is accompanied by the raw excerpts from the lecture or book on which it was based,"explains our student.

MiNI team at the prestigious IJCAI-ECAI conference

The live demonstration attracted considerable interest from experienced researchers.

"What we liked most was the enthusiastic response to our solution during the demonstration session. Interest at the stand was so great that we had to divide our team into three groups and demonstrate the plug-in simultaneously on several laptops. New visitors were constantly approaching us to find out what we had created and to ask further questions. We noted those questions and will focus on them if we continue developing our solution," emphasises Michał Kukla. "We came away from the conference with very positive impressions. It was our first major international event of this kind. Having our undergraduate engineering projects accepted at such a highly ranked conference was a tremendous distinction. We gained first-hand experience of explaining technological complexities clearly and discussing our solution with experts from around the world. Those conversations gave us a wealth of ideas for the system’s further development. By attending other talks, we were also able to learn first-hand about the latest trends in AI."

Our team’s paper is available here: https://arxiv.org/abs/2605.06963.