University of Luxembourg - Empowering Education with AI

Prof. Eric Tschirhart: At the University of Luxembourg, we had been looking at large languagemodels for a while, but what I saw was mostly hype. My students were even paying for tools like

ChatGPT, only to receive long, unfocused, and sometimes contradictory answers.I had this idea some time ago of creating an agent that could use my course materials to assiststudents, specifically in understanding and responding to the course's learning objectives.

‍

In myfield, this is relatively straightforward because the curriculum is regulated by a European directivethat standardizes the learning objectives.That means we can’t just teach anything; we must adhere to the defined framework. This addsmore complexity: when designing or guiding teaching and learning content, we must also ensurecompliance with these regulations.Our challenge was finding an AI solution that was educationally sound, GDPR-compliant, andinclusive.

Dr. Christian Grévisse: When we looked at the AI tools already on the market, many of themseemed appealing at first glance. Some learning management systems (LMSs) even include AIfeatures. However, the problem was always the same: they worked alongside the teacher, ratherthan with the teacher. You couldn’t feed them your own material in a structured way, you couldn’tset guardrails, and you had no real control over the answers they gave. For me, that wasunacceptable. I needed something that would only respond within the boundaries of my courseobjectives and materials.The EDT&Partners team made the entire process simple for us, giving us the benefits of AWSwithout the headache. In the end, it wasn’t just about choosing a tool—it was about selecting apartner focused on customer needs who could help us build exactly what we needed and guideus through the entire process.What convinced me even more was how the EDT&Partners team worked with us. They wereresponsive, they listened carefully, and they weren’t afraid to ask the hard questions that made theproduct better. We had weekly meetings, sometimes twice a week, and most issues were resolvedin a day or two. This level of responsiveness is rare in academia.

Eric Tschirhart: The primary goal was to amplify the impact, data, and experience of lectures byleveraging advanced AI capabilities. The implementation helps us enhance faculty lecturepreparation through intelligent content analysis and support tools. It also contributes to studentengagement and learning outcomes with personalized, AI-driven resources. Additionally, we nowhave access to comprehensive analytics that enable us to measure and optimize lectureeffectiveness. For us, the implementation of Lecture happened in two main phases. The firstphase was the development phase, during which I shared all my course materials and pedagogicalcontent with the EDT&Partners team.Dr. Christian Grévisse: The second phase, on the technical side, was actually quite simple. Thecontact with the tech team was steady and always responsive. If there were any bugs, they werefixed quickly. For any LMS administrator, it’s really straightforward. The documentation iscomprehensive, containing everything you need, and the process itself is straightforward

Prof. Eric Tschirhart: For me as a teacher, Lecture has forced me to raise the quality of mycontent. Preparing material for an AI agent means that every slide and every explanation must beprecise. It takes time, yes, but it also makes the course stronger. Instructors at the university canutilize generative AI to create personalized learning paths, evaluate assessments, translatedocuments, and generate creative assets. Another benefit is assessment. Because I feed my ownexam questions into Lecture, students can practice with the kinds of problems I actually expectthem to solve. It aligns their preparation with what really matters.

Prof. Eric Tschirhart: At the University of Luxembourg, we had been looking at large languagemodels for a while, but what I saw was mostly hype. My students were even paying for tools like

ChatGPT, only to receive long, unfocused, and sometimes contradictory answers.I had this idea some time ago of creating an agent that could use my course materials to assiststudents, specifically in understanding and responding to the course's learning objectives.

‍

In myfield, this is relatively straightforward because the curriculum is regulated by a European directivethat standardizes the learning objectives.That means we can’t just teach anything; we must adhere to the defined framework. This addsmore complexity: when designing or guiding teaching and learning content, we must also ensurecompliance with these regulations.Our challenge was finding an AI solution that was educationally sound, GDPR-compliant, andinclusive.

Prof. Eric Tschirhart: At the University of Luxembourg, we had been looking at large languagemodels for a while, but what I saw was mostly hype. My students were even paying for tools like

ChatGPT, only to receive long, unfocused, and sometimes contradictory answers.I had this idea some time ago of creating an agent that could use my course materials to assiststudents, specifically in understanding and responding to the course's learning objectives.

‍

In myfield, this is relatively straightforward because the curriculum is regulated by a European directivethat standardizes the learning objectives.That means we can’t just teach anything; we must adhere to the defined framework. This addsmore complexity: when designing or guiding teaching and learning content, we must also ensurecompliance with these regulations.Our challenge was finding an AI solution that was educationally sound, GDPR-compliant, andinclusive.