Collaborative Learning with Generative AI: A Dual-Layer Metacognitive Scaffolding Framework for Developing Hybrid Intelligence

• In Event: AERA Poster Session Sunday 11:45 am
  In Poster Session: Technology, Instruction, Cognition, and Learning SIG Poster Session

Abstract

This study proposes a Dual-Layer Metacognitive Scaffolding (DLMS) framework, which aims to enhance learners’ self-regulated learning and interaction with Generative AI (GAI) in a hybrid intelligence environment. The framework was implemented through CoLearn.AI—a DLMS-supported pedagogical agent (DLMS-PA) powered by a large language model. Eight Year-2 university students majoring in English language education in Hong Kong participated in the study by designing lesson plans, sequentially supported by standard GAI and CoLearn.AI. Lesson plans and participant-GAI interaction logs were analyzed using expert ratings and Epistemic Network Analysis. Findings indicate that CoLearn.AI promoted more coherent collaboration and organized learning behaviors, leading to higher-quality lesson designs. This study contributes to the design of theory-informed scaffolding approaches that promote meaningful and productive learner-GAI partnerships.

Authors

• Miao Yue, Chinese University of Hong Kong

• Lillian Wenxuan Liu, University of California - Irvine

• Lap Kwan Alex LAM, The Education University of Hong Kong

• Morris Siu-Yung Jong, Chinese University of Hong Kong