From Attitude to Action: The Mediating Role of AI-Supported Scaffolding in Middle School Mathematics Learning

Abstract

Despite widespread artificial intelligence (AI) integration in mathematics education, the mechanisms through which student engagement translates into adoption behavior remain undertheorized. Existing technology acceptance frameworks overlook the pedagogical dimension of learner-AI interaction. Grounded in Vygotsky's Zone of Proximal Development (ZPD), this study tests agentic scaffolding perception as a behavioral mediator in middle school mathematics. A correlational mediation design was employed with 578 students (Grades 5–8; Türkiye; 2025–2026) using the 14-item AS-AISM (CFI = .973; RMSEA = .036; α = .851) and Hayes's PROCESS Macro (Model 4) with bootstrapped confidence intervals. Emotional–cognitive engagement predicted both the perceived instructional role of AI (b = 0.65, p < .001) and reported use of AI tools in studying (b = 0.66, p < .001). The perceived instructional role of AI was a statistically significant but small mediator (indirect = 0.057; 95% CI [0.003, 0.115]), accounting for 8.7% of the total effect, while the direct pathway remained dominant (b = 0.60, p < .001). Pedagogical framing therefore contributes to adoption, but attitude translates into study behaviour largely through a direct route. Findings extend ZPD theory into AI-mediated mathematics learning.