Session T-07 - Topics: AI‑Mediated Learning; Agentic Framework Studies; Scientific-computing Platform Development

This presentation reports on a five‑year systematic review of empirical studies examining how AI agents support collaborative learning in higher‑education STEM. As part of a larger LAVIA‑funded study, we apply a four-level framework (Yan, 2025) — Adaptive Instrument to Peer Collaborator—to classify agent roles, with emphasis on co‑learning and peer‑like collaboration. Preliminary findings show few high‑agency implementations, revealing gaps and design opportunities. We conclude with implications and next steps: developing AI agents that support collaborative inquiry and shared knowledge construction.

The SCripting and ORchestration Environment (SCORE) offers a technology infrastructure for active learning with a theoretical emphasis on learning communities. SCORE supports the design and enactment of collective forms of inquiry for small and large courses, allowing flexible and dynamic student grouping based on real-time analysis of student responses. It supports the formation and dynamic use of a community knowledge base, and new forms of learning analytics and assessment. We present studies of small and large courses and a makerspace.

There is an increasing demand for college-level science students with more advanced computational skills. An ideal environment for supporting such learning would be an online scientific-computing platform tailored to their specific pedagogical needs. Until the recent emergence of AI coding agents, such a platform would never have been considered due to its high cost in design, development and maintenance. Having undertaken such a project over the last 2 years, we share with the community our experiences and best practices.