Gaia Multi-agent Tutor at AIED Conference
We are delighted to share an important milestone for the L2BGreen project: our first scientific publication has been accepted. The paper “Learning to Be Green: A Pilot Study of the Gaia Multi-agent Tutor” will be presented at the International Conference on Artificial Intelligence in Education (AIED). This marks a significant step for the project and highlights the scientific relevance of our work at the intersection of artificial intelligence, sustainability education and green entrepreneurship.
The publication, authored by Ahsen Çini, Pantelis M. Papadopoulos, Stergios Tegos and Apostolos Mavridis, presents a pilot study of Gaia, a multi-agent conversational tutor designed to support learners in developing a green entrepreneurial mindset.
Exploring AI-supported learning for green entrepreneurship
Gaia was developed to guide learners through case-based learning activities in the field of green entrepreneurship. The tutor supports students as they move from understanding case contexts to deeper analysis, argumentation and application.
The pedagogical design combines Case-Based Learning, Bloom’s Taxonomy, Socratic scaffolding and AI-supported feedback. This allows Gaia to act not simply as a chatbot, but as a structured learning companion that supports deeper engagement with complex sustainability-related business challenges.
Author statement by Ahsen Çini
“With Gaia, we wanted to explore how conversational AI can support students not only in accessing information, but in actively reasoning through sustainability-related business cases. The pilot study shows promising signs that students are willing to engage deeply with an AI tutor when the interaction is pedagogically structured.”
A theory-driven multi-agent tutor
A central contribution of the paper is Gaia’s multi-agent architecture. The system is designed as an “Interaction Engine” with different agents responsible for specific pedagogical functions. These include presenting case scenarios, scaffolding knowledge, supporting application and providing feedback on argumentation.
This structure reflects one of the key principles of L2BGreen: AI in education should not be used as a shortcut, but as a carefully designed support system that helps learners think more critically, apply knowledge and reflect on complex real-world challenges.
Author statement by Pantelis M. Papadopoulos
“The study underlines the importance of grounding AI-based learning tools in educational theory. Gaia was designed around clear pedagogical principles, ensuring that students are guided through different levels of understanding, analysis and application rather than simply receiving quick answers.”
Promising insights from the pilot study
The pilot study involved 25 undergraduate students who interacted with Gaia over three weeks across three online case studies. These cases addressed topics such as money impact, greenwashing and design thinking. The analysis focused on system logs and post-activity evaluations.
The results showed consistently high engagement. Students spent more than 45 minutes per case on average and sent more than 25 messages per week to Gaia. The number of student messages increased significantly from the first to the third case, suggesting growing interaction over time.
Students also reported positive perceptions of Gaia’s usability and the quality of the learning materials. The study further found that students with lower prior orientation toward green entrepreneurship tended to have longer discussions with Gaia, indicating that the tutor may be especially valuable for learners who need more support in building understanding.
Author statement by Stergios Tegos
“One of the most encouraging findings is that learners did not only complete the activities, but actively interacted with Gaia. The increasing number of student messages across the cases suggests that a well-designed conversational tutor can stimulate meaningful engagement over time.”
Learning from student feedback
The study also highlights valuable feedback from learners. Students appreciated Gaia’s ability to clarify misunderstandings, rephrase questions and support their answers. At the same time, some participants found the text-based format tiring and suggested shorter, more focused dialogues.
This feedback is highly relevant for the next development steps of L2BGreen. It points toward the potential of future multimodal extensions and more refined interaction formats that can make AI-supported learning even more accessible, engaging and learner-friendly.
Author statement by Apostolos Mavridis
“The pilot study gives us valuable technical and pedagogical insights for the further development of Gaia. Student feedback shows the potential of the system, while also helping us identify where interaction design, usability and future multimodal features can be improved.”
A shared achievement across the L2BGreen consortium
This first publication is an important milestone for the entire L2BGreen consortium. It reflects the collaborative effort of partners working across educational technology, artificial intelligence, sustainability, entrepreneurship and instructional design.
Looking ahead
The presentation at AIED offers an excellent opportunity to share L2BGreen’s work with the international AI in education community and to discuss how pedagogically grounded AI systems can contribute to meaningful, responsible and future-oriented learning experiences.
For L2BGreen, this first publication is more than a research output. It is a milestone that reflects the project’s growing impact and the commitment of all partners to advancing education for a more sustainable future.
