A deep learning-integrated RADEC model for higher-order thinking in preservice elementary teachers’ environmental science learning

In this study, deep learning (DL) refers to a pedagogical approach—not an artificial intelligence or machine learning technique—centered on meaningful, mindful, and joyful engagement with content (Kemendikdasmen, 2025). Higher-order thinking skills (HOTS)—analyzing, evaluating, and creating—are essential for preservice elementary teachers, who must develop HOTS themselves before they can teach it to pupils, yet HOTS remain underdeveloped in early teacher-education coursework. This study tested whether integrating Indonesia’s DL principles (meaningful, mindful, and joyful learning) into the read-answer-discuss-explain-create (RADEC) model improves HOTS more than direct instruction. Using a true experimental, pre-/post-test control-group design with random assignment to condition, 172 third-semester primary school teacher education students at a public university in Indonesia (86 experimental, 86 control) completed a unit on environmental problems. HOTS was measured with a 20-item multiple-choice test spanning the analyze, evaluate, and create dimensions (Cronbach’s alpha = .859). The experimental group’s post-test scores (mean [M] = 77.56, standard deviation [SD] = 9.904) were significantly higher than the control group’s (M = 69.42, SD = 10.72), t(170) = 5.17, p < .001, d = 0.79, with descriptive gains concentrated in evaluate (+16.94 points) and create (+12.21 points), suggesting where RADEC’s discuss-explain-create sequence may contribute most. A one-way multivariate analysis of covariance confirmed that these dimension-level gains were statistically significant after controlling for pre-test scores, Wilks’ lambda = .82, F(3, 165) = 11.77, p < .001, partial η² = .18, with significant univariate effects for analyze, evaluate, and create considered separately.