The rapid integration of generative AI (GenAI) into immersive learning technologies is opening new possibilities for adaptive and embodied learning. Mixed reality (MR) environments, when combined with embodied AI agents, offer particular promise for collaborative learning by blending physical–virtual interactions and enabling real-time, intelligent support. Yet few studies have examined how to design and implement MR environments, and existing work often lacks strong theoretical grounding. This study introduces Embodyverse, an MR environment with an embodied AI robot to support socially shared regulation of learning (SSRL) for collaborative learning. Drawing on SSRL theory and principles of adaptive support, regulation is conceptualised as a cyclical process involving task preparation, goal setting and planning, performance, and reflection and adaptation. Technically, Embodyverse integrates MR head-mounted displays, an embodied AI robot powered by vision-language models and large language models (LLMs), and a layered architecture for perception, cognition, and interaction. The AI robot provides adaptive, context-sensitive support through timely, triggered interactions with learner groups. The framework is demonstrated through an inquiry-based plant growth scenario on seed germination, where learners collaboratively manipulate environmental conditions and observe biological processes. This study offers a theoretically grounded design framework for Embodyverse and provides insights from analysing participants’ feedback and advice during the pilot implementation, to advance understanding of SSRL within MR-based collaborative learning environments.
Song et al. (Tue,) studied this question.