Pui Ching Primary School, Ho Man Tin, Kowloon, Hong Kong SAR, China

Abstract: This paper presents the design, development, and evaluation of an AI-Enhanced Adaptive Diagnostic Assessment and Scaffolding (AI-ADAS) system for primary mathematics education, developed through a design-based research approach in a Hong Kong primary school. Addressing the persistent challenge of providing timely, personalized diagnostic assessment and differentiated instruction in large mathematics classes, the AI-ADAS system integrates artificial intelligence to create a dynamic assessment loop that continuously diagnoses student understanding, provides adaptive scaffolding, and informs teachers' instructional decisions. The system features a three-layer architecture: (1) an AI-powered diagnostic assessment engine that identifies specific misconceptions and knowledge gaps across the four mathematics strands (Number, Measures, Shape & Space, Data Handling); (2) an adaptive scaffolding mechanism that provides graduated support matched to each student's zone of proximal development; and (3) a teacher dashboard that visualizes learning analytics and supports data-informed instructional planning. Conducted over three iterative design cycles with Primary 4 and 5 students (n=126) and four mathematics teachers at Pui Ching Primary School, the study traces the evolution of the AI-ADAS system from initial prototype to refined implementation. Findings indicate that the system significantly improved students' mathematics achievement (F=12.35, p<.001, η²=.17), particularly for students with learning difficulties, and enhanced teachers' capacity for diagnostic assessment and differentiated instruction. Qualitative data further reveal that teachers valued the system's ability to pinpoint specific misconceptions and free up time for more targeted instructional support, while students appreciated the immediate, personalized feedback. The study contributes to the growing body of knowledge on AI-enhanced assessment in mathematics education and offers practical insights for Hong Kong educators seeking to implement assessment for learning through technology integration.

Keywords: AI in education, adaptive assessment, diagnostic assessment, mathematics education, primary school, scaffolding, design-based research, Hong Kong education