Deep Learning-Based STEM: A Time Series Experimental Study of Force Problem-Solving Skills Yunisa Sapphira Titalia & Dede Trie Kurniawan
Universitas Pendidikan Indonesia
Abstract
Problem solving is one of the essential 21st-century skills. Previous research indicates that elementary school students^ problem-solving skills remain low, largely due to learning practices that do not adequately support their development. The STEM approach is considered effective in enhancing these skills, especially when integrated with deep learning to create a meaningful and interdisciplinary learning environment. Therefore, this study aims to determine the effect of a STEM approach based on deep learning on students^ problem-solving skills, as well as to measure the extent of improvement resulting from the treatment. This research employed a time series design with one pretest and three posttests. The participants were 28 fourth-grade elementary school students. The findings revealed that the STEM approach based on deep learning had a significant effect on students^ problem-solving skills. Moreover, the average improvement across the three N-Gain scores was categorized as moderate and quite effective. Furthermore, misconceptions in spring force and gravitational force decreased more consistently after the treatment compared to other force concepts.
