AI-based scaffolding and conceptual understanding: evidence from Indonesian students using PLS-SEM

Abstract

The rapid integration of artificial intelligence (AI) in higher education has increased interest in AI-based scaffolding (AIS) to support conceptual learning, particularly in teacher education. However, empirical evidence explaining how learners’ cognitive, self-regulatory, and technological characteristics jointly shape perceptions of AIS remains limited, especially in developing country contexts. This study examines the predictive relationships among conceptual understanding (CU), cognitive engagement (CE), self-regulated learning (SRL), perceived ease of use (PEOU), AI self-efficacy (AISE), and perceived scaffolding quality in explaining Indonesian undergraduate teacher education students’ perceptions of AIS. Using a quantitative explanatory–predictive design, data were collected from 157 students and analyzed using partial least squares structural equation modeling (PLS-SEM). The results indicate that SRL, PEOU, and AISE are the strongest predictors of perceived AIS, while CU, CE, and scaffolding quality also show significant positive associations. These findings highlight the importance of learner readiness and instructional design in AI-enhanced learning environments. Practically, teacher education programs should integrate AI scaffolding that explicitly supports self-regulation, builds students’ confidence in using AI tools, and promotes sustained CE in complex learning tasks.