Clustering-Based AI Model for Hypothetical Learning Trajectories to Improve Student Metacognitive Skills
Year: 2026
Type: preprint
Study design: Clinical Trial
Abstract: This study developed and validated an AI-based Hypothetical Learning Trajectory (HLT) framework to foster metacognitive skills among 72 secondary school students over an eight-session adaptive learning intervention. Drawing on Flavell's (1979) metacognitive theory and Simon's (2020) HLT model, the research employed K-Means clustering to identify distinct learner profiles based on twelve behavioral and metacognitive features extracted from a digital learning platform. Clustering analysis revealed three meaningful profiles: low metacognitive (33%), moderate metacognitive (39%), and high metacognitive (28% of students). These profiles were then translated into differentiated HLT pathways—structured scaffolding for low learners, monitoring feedback with strategy adjustment for moderate learners, and self-directed inquiry with meta-evaluation for high learners. Longitudinal assessment measured metacognitive gains across eight sessions using pre-post comparisons and learning rate analysis. Results demonstrated that all clusters improved significantly, with low metacognitive learners showing the largest relative gain (88% performance increase) and moderate learners achieving the largest absolute gain (+35 points). However, the achievement gap between low and high groups persisted, confirming Stanovich's (1986) Matthew effect in metacognitive development. PCA revealed that a single latent component explained 92.8% of behavioral variance, indicating that metacognitive competence functions as a unified continuum in digital trace data. The study contributes empirical validation that clustering-derived HLT pathways can be successfully operationalized within AI-mediated adaptive systems. Implications for instructional design, AI system development, and educational policy are discussed, alongside limitations regarding sample size, control group absence, and measurement validity.
Source: Research Square
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
SDG: Quality education
Open Access status: green
OpenAlex ID: W7156309037
