AI-Assisted Case Method Learning for Scientific Writing: A Design and Development Study

Abstract

work, yet writing instruction remains largely lecture-driven and disconnected from students’ digital tools. This study designed, validated, and tested instructional products for an undergraduate Scientific Writing course that integrate the case method with Perplexity.ai, an AI answer engine providing source-linked responses. Combining an adapted Recursive Reflective Design and Development (R2D2) model with the Research-Development-Research (RDR) cycle, the study produced lesson plans, thirteen teaching-material units, a four-stage case-method syntax with explicit operational and ethical rules for AI use, and authentic assessment instruments, including a 22-indicator writing rubric (0–100 scale). Three practitioners and three experts judged all five products valid (M = 3.40 on a 4-point scale), and their feedback guided revisions through small-group and large-group trials and a semester-long implementation at Universitas Negeri Padang. Both the model class (n = 21; M = 65.24 to 83.33) and a conventionally taught class (n = 18; M = 62.22 to 72.78) improved significantly (Wilcoxon p ≤ .001). With classes equivalent at pre-test, analysis of covariance confirmed the model class’s advantage, F(1, 36) = 14.19, p = .001, partial η² = .28, adjusted difference 8.86 points, corroborated non-parametrically (p = .005). Students reported high knowledge gains (90.5%) and positive attitudes (81.0%). The findings establish the products’ validity and implementability, provide non-randomised evidence that the model outperformed conventional teaching, and show how responsible AI use can become an explicit object of writing instruction.