Article
Development of an AI Chatbot-Integrated Physics E-Module Based on Deep Learning Pedagogy with Adaptive Scaffolding to Enhance High School Students’ Conceptual Understanding
Jurnal Amal Pendidikan, Vol. 7 No. 2 (2026), hlm. 239–253
Abstrak
Conceptual difficulties in school physics require structured learning activities and responsive support. We developed Fisika.AI, a web-based e-module integrating guided inquiry, virtual experimentation, and an AI chatbot for adaptive scaffolding. Deep-learning pedagogy denotes meaningful, reflective conceptual learning. An ADDIE development study incorporated a one-group pretest-posttest evaluation in two Indonesian public senior high schools. Observations, interviews, and questionnaires informed needs analysis. Product validity was assessed through expert agreement, practicality through two teachers and 72 Grade XI students, and conceptual change through 72 complete paired SOLO records. Gregory coefficients were 0.840 for content, 0.875 for media, and 0.800 for instructional design. Teacher and student practicality reached 90.00% and 89.10%. Internal consistency was limited: student practicality α = .127; 25-subquestion conceptual test α = .403 at pretest and .494 at posttest. Mean conceptual scores increased from 1.737 to 3.567 (N-gain = .561; t(71) = 73.707, p < .001). At posttest, 71 students were Multistructural and one Relational under the study-specific score bands. These results document positive appraisals and observed score changes. The uncontrolled design, measurement limitations, and unresolved questionnaire chronology preclude causal attribution or isolation of the chatbot contribution. Relational reasoning and transfer remain development priorities.