Speaker
Abstract section 3: Content/method
educational platform Minecraft Education Edition. Students complete structured construction challenges, such as building environmentally sustainable structures within defined constraints. AI agents provide guidance through rule-based demonstrations, contextual hints, knowledge bases, and automated feedback. Some experimental conditions include AI assistance, while others do not, allowing comparisons of task completion, design quality, and interaction patterns. Data collected includes building outcomes, task completion times, and student interaction with AI guidance.
Abstract section 4: Outcomes/results
Preliminary experiments suggest that AI agents can function as effective scaffolding tools in spatial learning tasks, particularly when providing demonstrations and contextual feedback. At the same time, limitations emerge in interpreting complex design choices and collaborative behavior among participants. These findings highlight both the potential and constraints of deploying AI systems in interactive environments beyond text. The presentation will discuss implications for the development of AI-supported learning systems and methods in virtual worlds and other multimodal digital environments in the classroom.
Abstract section 5: References
Liu, S., Li, Y., Zhang, K., Cui, Z., Fang, W., Zheng, Y., ... & Song, M. (2024). Odyssey: Empowering minecraft agents with open-world skills. arXiv preprint arXiv:2407.15325.
Matsushita, N., Mizoi, S., & Saito, D. (2024). A survey of perceptions in AI education using Minecraft and generative AI. In Proceedings of the 14th Annual Conference of the Digital Games Research Association Japan (pp. 241–245). Digital Games Research Association Japan.
Teixeira, D. (2025). Mining for Education: A Literature Review of Minecraft Used in Educational Environments. Teikyo Daigaku Gaikokugo Gaikoku Bunka, 16, 181-209.
Abstract section 1: Relevance
Much current discussion of artificial intelligence focuses on language models and text-based interaction. However, educational environments are increasingly more interactive and multi-sensory, involving visual, spatial, and interactive media. Virtual worlds present an opportunity to investigate how AI may operate in non-text environments where learning occurs through building, experimentation, and collaboration. Investigating AI in such environments contributes to understanding how AI may support experiential learning and design-based pedagogy in free-form environments. My research addresses broader questions about the role of AI in multimodal task-based learning environments and how AI can interact with human activity beyond textual communication or analysis.
Abstract section 2: Contribution/research questions
What forms of interaction emerge when AI operates in spatial, construction-based learning tasks?
What limitations appear when AI systems are applied to non-text interactive environments?
How effectively can AI agents assist learners in task-based activities in a virtual sandbox environment?
| Title | AI Beyond Text: Experiments in Minecraft Learning Worlds |
|---|---|
| Commercial Promotion | No |
| Language | English |
| Teaching Context | General |