Informed Undergraduate Teaching Reform in the Building Materials Course Under New Agricultural Science: Based on the “One Core, Two Channels, and Four-Dimensional Drivers” Model

Artificial intelligence (AI) offers new opportunities for evidence retrieval, data interpretation, experimentation, feedback, and evidence-based problem solving in engineering education. This study developed a theoretically grounded “One Core, Two Channels, and Four-Dimensional Drivers” model to align AI, research-informed teaching, and New Agricultural Science within an undergraduate Building Materials course, and examined preliminary between-cohort outcome differences. A 32-teaching-hour nonequivalent comparison-group study involving 123 students was conducted. The intervention integrated research-informed cases, rural and green-construction contexts, purpose-specific AI-supported tasks, laboratory work, staged projects, and multidimensional assessment. The intervention cohort scored higher on the final examination (mean difference 8.00 points; Hedges’ g = 1.19), experimental grade (4.67; g = 1.28), continuous-assessment score (35.56; g = 3.34), and course evaluation (1.63 on a 70-point scale; g = 0.58). Project innovation data were highly sparse. The continuous-assessment difference should not be interpreted as a standalone AI literacy effect because structured AI-learning opportunities differed by condition, and equivalence of non-examination scoring was not independently established. Given nonrandomized allocation, the absence of participant-level baseline measures, and unequal measurement strength across outcomes, the findings represent preliminary between-cohort evidence rather than causal treatment effects.