The Impact of the GenAI-Based Reading Model (GBR) on Students' Visualization of the Greenhouse Effect in Primary Education

This study aims to examine the effect of the GenAI-Based Reading (GBR) learning model on elementary school students' visual representations in understanding the concept of the greenhouse effect. Visual representation, as an integral part of scientific literacy, plays a crucial role in helping students build mental models of complex scientific phenomena. In the context of elementary education, the ability to visually represent scientific concepts is a challenge, particularly for abstract topics such as the greenhouse effect. This study used a mixed-methods approach with a multiphase explanatory design, involving 28 fourth-grade students at a public elementary school in West Java. The main instrument was a Three-Tier Test that measured students' conceptual understanding, confidence level, and visual representation skills. Visual representations were analyzed using a six-category rubric: Scientific Drawing, Partial Drawing, Undefined Drawing, Misconception Drawing, Non-Microscopic Drawing, and No Drawing. The analysis showed that GBR-based learning had a positive impact on improving students' visual representations, indicated by the elimination of misconceptions and an increase in the number of students in more meaningful representation categories. Statistical analysis using the Wilcoxon signed-rank test showed a significant difference between the pretest and posttest (p = 0.0256). However, most students remained in the initial category, indicating the need for further pedagogical intervention. Qualitative analysis revealed that students' representations were still dominated by macroscopic elements and surface symbols, not fully reflecting sub-microscopic scientific mechanisms. Factors such as the limited duration of the intervention, the lack of visual scaffolding, and low student confidence also influenced the results. The teacher's role as a conceptual facilitator and fosterer of epistemic courage is key to optimizing the potential of AI technology in science learning. This study concludes that the GBR model is effective as an initial catalyst for developing scientific visual representations. However, its long-term success depends heavily on integrating the technology with a reflective and contextual pedagogical approach.