Activities
AI-Powered Cognitive Pattern Analysis for Personalized Learning
This project combines graph machine learning with large language models to analyze students’ exam responses. Its objective is to identify cognitive patterns indicative of how students comprehend and process information. The overarching aim is to utilize these insights for personalized educational strategies, enhancing learning outcomes by adapting to the distinct cognitive profiles of each student.
Researchers: Thiago Brant (LISER), Jun Pang (Uni.lu)
Student assistant(s): Julien Kühn (Uni.lu)
Chatwise: ChatGPT: as a High School Academic Tool for Writing, Innovation, Skills, and Education
The project aimed to equip both students and teachers with the knowledge to use ChatGPT responsibly and ethically, emphasizing the importance of crafting effective prompts, creating a platform for students and teachers feedback and discussions about the transformative potential of AI in personalized education. The project primarily focused on two key objectives: conducting workshops and hosting exhibitions at various events.
Researcher(s): Sana Nouzri
Students: Othmane Mahfoud, Parsa Vares
Teaching experiences using the example of the ‘Information Management 1’ course: to what extent can traditional learning be usefully combined with AI?
The study examines the extent to which traditional learning approaches can be combined with the growing importance of AI, as the integration of AI and human support does not have to be an either/or proposition. Rather, a combination is conceivable in which the advantages of both approaches can be utilised. AI could provide basic, automated support on the one hand, while the human instructor focuses on more in-depth mentoring and the promotion of critical thinking on the other. This combination could offer students both access to modern technology and the benefits of human, individualised support. This will be analysed using the example of student support in the Information Management 1 course.
Researcher(s): Prof. Dr. Martin Theobald
Student(s): Valentina Rondinelli
The role of AI assistants in the teaching of formal methods
The aim of the project is to understand to what extend Artificial Intelligent (AI) assistants help students to acquire knowledge and develop skills during the learning process.
The project is focused on a particular topic of the computer science domain referred to as Formal Methods (FMs). This topic strongly relies on mathematics; thus, it is hypothesised that current AI assistants are of marginal help to provide correct solutions (due to the underlying principles on which these AI assistants are built) but may help during the path towards the solution: this is the key point on which the project will try to shed light.
Researcher(s): Alfredo Capozucca
Student(s): Daniil Yampolskyi
Publications
Computing Education in the Age of AI-Based Assistants: Challenges and Opportunities
On the basis of discussions at the 2𝑛𝑑 edition of the Frontiers on Software Engineering Education workshop, researchers identified challenges brought by the use of AI assistants into computing education. These challenges represent a starting point for the endless road towards effective education of software engineering and computing science in higher education. This paper summarises the challenges and research opportunities that were identified during the heated discussions at the workshop.
Authors: Alfredo Capozucca, Sophie Ebersold, Jean-Michel Bruel & Bertrand Meyer
Fine-Tuning a Large Language Model with Reinforcement Learning for Educational Question Generation
Educational Natural Language Generation (EduQG) aims to automatically generate educational questions from textual content, which is crucial for the expansion of online education. Prior research in EduQG has predominantly relied on cross-entropy loss for training, which can lead to issues such as exposure bias and inconsistencies between training and testing metrics. To mitigate this issue, we propose a reinforcement learning (RL) based large language model (LLM) for educational question generation. In particular, we fine-tune the Google FLAN-T5 model using a mixed objective function that combines cross-entropy and RL losses to ensure the generation of questions that are syntactically and semantically accurate. The experimental results on the SciQ question generation dataset show that the proposed method is competitive with current state-of-the-art systems in terms of predictive performance and linguistic quality.
Authors: Salima Lamsiyah, Abdelkader El Mahdaouy, Aria Nourbakhsh & Christoph Schommer
Personalized learning with ChatGPT
The project investigates the enhancement of personalized learning through autonomous AI agents within a custom tutoring platform. The scientific deliverable provides a comprehensive overview of personalized learning, adaptive learning techniques, and agent-oriented programming (AOP). It explains how these concepts integrate to create tailored educational experiences, backed by key literature emphasizing the transformative potential of AI in education. The technical deliverable focuses on developing a multi-agent system using the CrewAI framework, implementing backend services with Django and SQLite, and creating a user-friendly frontend with React. The project demonstrates that integrating AOP and AI significantly enhances student engagement, motivation, and learning outcomes, providing a flexible and scalable educational environment tailored to individual learner needs. This work lays the groundwork for future advancements in AI-driven personalized learning systems, showcasing their potential to revolutionize traditional educational methods..
Supervisor(s): Sana Nouzri
Student(s): Hedi Tebourbi
Predicting cognitive errors in virtual reality by monitoring changes in pupil dilation during cognitive processing
As part of its broader research agenda, the VR/AR Lab explores the potential of AI technologies in educational mediated-reality applications. A recent paper demonstrated the possibility of predicting cognitive errors in virtual reality by monitoring changes in pupil dilation during cognitive processing. Applying machine learning methods to eye-tracking data, among others, has important implications for self-adaptive virtual environments and user interfaces in education and training.
Authors: Sahar Niknam, Jean Botev.
Über den Einsatz von Generative AI für die Lehre
The authors believe that generative AI is relevant when used in a measured, controlled manner, and as a relaxing element. They advise against completely replacing traditional teaching concepts and methods with generative AI, but suggest gradually and strategically introducing systems like ChatGPT as an additional tool.
Authors: Christoph Schommer, Salima Lamsyah, Sana Nouzri
Conferences
AI for Education
Workshop at the Zhejiang-Europe Center for Advanced Intelligence and Ethics
13-14 June 2024
Zhejiang University, Hangzhou
Speakers:
- Christoph Schommer: “AI for Education and Training!”
- Davide Liga: “LLMs: Stochastic Parrots or Oracles”
- Salima Lamsiyah: “Education in the Age of Generative AI: Leveraging Large Language Models and Reinforcement Learning for Educational Question Generation”