Tracking Learning Process on The ICE Model in A Flipped Classroom at A Japanese University

Abstract

In this paper, we present the results of an analysis of text data written on digital MANDALAs by students in the course on the fundamentals of the Universe offered at a Japanese university. We designed the course based on the ICE model and implemented it in a flipped classroom. We identified students' learning processes in I, C, or E by verbs and analyzed them with KH Coder. We observed statistically that students' learning transitioned from I to C or E in their thoughts from the first week to the last week of the course. To track the learning process more accurately from the context, we fine-tuned BERT with classified data in I, C, and E, and achieved a classification accuracy of 78%.