Schools are meant to prepare learners for the future outside of school. Current developments
in AI, machine learning and robotics suggests a future of fully shared social spaces, including
learning environments (LEs), with robotic personalities. Today’s learners (as well as teachers)
should be prepared for such a future. AI in Education (AIED) has focused on implementation
of online and screen-based Pedagogical Agents (PAs); however, research findings support
richer learning experiences with embodied PAs, hence, recent studies in AIED have focused
on robot as peers, teaching assistants or as instructional materials. Their classroom uses
employ gamification approaches and are mostly based on a one-robot- one-student interaction
style whereas current educational demands support collaborative approaches to learning.
Robots as instructors are novel, considered a major challenge due to the requirements for
good teaching, including the demands for agency, affective capabilities and classroom control
which machines are believed to be incapable of. Current technological capabilities suggest a
future with full-fledged robot teachers teaching actual classroom subjects, hence, we
implement a robot teacher with capabilities for agency, social interaction and classroom
control within a collaborative learning scenario involving multiple human learners and the
teaching of basic Chemistry in line with current focus on STEM areas. We consider the PI
pedagogical approach an adequate technique for implementing robotic teaching based on its
design with inherent support for instructional scaffolding, learner control, conceptual
understanding and learning by teaching. We are exploring these features in addition to the
agentic capabilities of the robot and the effects on learner agency as well as improved
learning in terms of engagement, learner control and social learning. In the future, we will
focus on other key concepts in learning (e.g. assessment), other types of learners (e.g.
learners with cognitive/physical disabilities), interaction styles and LEs. We will also explore
and cross-community approaches that leverage on integration of sibling communities.