Educational technology research and development Education
From information delivery to cognitive co-creation: how generative AI enhances computational thinking education through student-authored visualizations
Abstract
Generative Artificial Intelligence (Generative AI) is increasingly used in computational thinking (CT) education, yet its role in enabling students to create and revise external representations of their understanding remains underexplored. The Generative AI-assisted cognitive authoring system (GACS) uses students’ everyday-language descriptions to generate interactive knowledge-visualization web pages that students can inspect and refine iteratively. Using a five-week quasi-experimental design involving two intact classes, the eighth-grade sample (N = 58) comprised 28 students in the GACS-supported practice (GSP) condition and 30 in the conventional representational practice (CRP) condition, with the two conditions sharing the same teacher-led instructional materials and differing in the practice segment of each lesson. After controlling for pre-test scores, adjusted means favored the GSP condition for overall self-reported CT disposition scores (3.37 vs. 3.15, false discovery rate (FDR)-adjusted p = .032, partial η^2 = .111), Learning Motivation scale scores (3.42 vs. 3.32, FDR-adjusted p = .037, partial η^2 = .097), and Metacognitive Awareness scale scores (3.34 vs. 3.28, FDR-adjusted p = .037, partial η^2 = .093). Algorithmic thinking showed the largest CT dimensional difference (partial η^2 = .116), whereas metacognitive evaluation did not reach statistical significance (FDR-adjusted p = .079, partial η^2 = .055). The uneven dimensional pattern suggests that explicit quality criteria combined with guided comparison and discussion may make evaluative standards more visible during cognitive authoring.