Fang, HsuanWen
Department of Computer Science, University of Taipei
Email：m11316009@go.utaipei.edu.tw
Lai, AhFur
Department of Computer Science, University of Taipei
Email：laiahfur@gmail.com
Chen, YenHung
Department of Computer Science, University of Taipei
Email：yhchen@utaipei.edu.tw
Abstract
In the era of rapid advancements in artificial intelligence, generative AI technologies such as ChatGPT and other large language models (LLMs) are increasingly being integrated into the field of education. This study explores the effectiveness and challenges of applying generative AI as a learning companion in upper elementary school mathematics instruction. Utilizing an action research methodology, this study was conducted in a sixth-grade mathematics unit on “quantitative relationships.” Instructional design incorporated two platforms: the personalized learning platform “Taiwan Adaptive Learning Platform” and the AI-based tool “Edo Learning Partner.” Students were guided to engage in self-directed learning through interactive dialogue with the AI learning partner. Findings reveal that generative AI can offer various forms of scaffolding to support differentiated instruction. However, the study also found that approximately 70% of students struggled to respond effectively during their conversations with the AI partner, which posed a significant learning bottleneck. Based on these results, it is recommended that teachers provide more detailed question prompts or exemplars, or integrate additional educational technologies to help students learn how to engage in meaningful “dialogue” with AI and receive effective feedback for learning.
Keywords: generative AI , learning companion , Primary school math problem solving , quantitative relationships, action research