AI-assisted personalized physical education for inclusive classrooms

Thesis Abstract

The present study addresses the persistent inequities in physical education (PE) experiences for students with diverse abilities by examining how AI-assisted personalized PE can enhance participation, engagement, and learning outcomes in inclusive classroom settings. Despite policy mandates for inclusive education, traditional PE often fails to accommodate individual differences in physical ability, motivation, and prior skill levels, leading to disengagement and suboptimal health trajectories. This research aims to investigate how AI-driven adaptive instruction, performance feedback, and activity tailoring can create equitable PE opportunities for all students. The primary objective is to evaluate the effectiveness of an AI-enabled PE platform in delivering individualized curricula that align with students’ motor competencies, preferences, and disability considerations, while maintaining core fitness and psychosocial objectives. Specific objectives include (i) determining whether AI personalization improves objective motor skill competence (MSA scores) and moderate-to-vigorous physical activity (MVPA) levels compared with standard PE; (ii) examining changes in student motivation, self-efficacy, and inclusion perceptions; and (iii) identifying teacher experiences, perceived feasibility, and classroom dynamics when integrating AI-assisted instruction. A mixed-methods design will be employed in a three-phase study conducted over an academic year. The quantitative strand will use a quasi-experimental design with two middle-school PE classes (n = 48 students; 24 per class), randomly assigned to AI-assisted personalized PE (n = 24) and conventional PE (n = 24). Data collection will include validated motor skill assessments (Test of Gross Motor Development–Third Edition), accelerometer-derived MVPA (taken during three 20-minute PE sessions per term), and standardized questionnaires measuring intrinsic motivation (Intrinsic Motivation Inventory), self-efficacy for physical activity (Children’s Self-Perception of Physical Activity Scale), and perceived inclusivity (Perceived Classroom Inclusion Scale). Data will be analyzed using multivariate analysis of covariance (MANCOVA) controlling for baseline scores, with post hoc pairwise comparisons and effect sizes reported. A longitudinal growth model will assess trajectories across terms. The qualitative strand will involve semi-structured interviews and focused group discussions with students (n = 12), PE teachers (n = 4), and inclusive education coordinators (n = 2). Thematic analysis will be guided by Braun and Clarke’s approach to identify experiences with AI personalization, perceived barriers and enablers, and impacts on classroom climate. Triangulation will integrate quantitative and qualitative findings to elucidate the mechanisms by which AI-driven adaptation influences learning and participation. The AI system will incorporate reinforcement learning to optimize task difficulty, activity choice, and feedback modalities based on real-time sensor data and periodic performance assessments. The theoretical framework combines Social Cognitive Theory (Bandura) to explain self-regulation, motivation, and observational learning with the Universal Design for Learning (UDL) principles to ensure multiple means of engagement, representation, and action/expression. Anticipated findings include (i) greater improvements in motor skill competence and higher MVPA engagement in the AI-assisted group; (ii) enhanced intrinsic motivation, self-efficacy, and sense of belonging among students with disabilities or lower initial skill levels; (iii) positive teacher perceptions regarding feasibility, adaptability, and time efficiency, accompanied by insights into professional development needs. The study will contribute to knowledge by providing robust evidence on how AI-enabled personalization can operationalize inclusive PE, align with UDL and equity objectives, and inform scalable implementation in diverse school contexts. Implications for practice include guidelines for selecting AI-augmented PE platforms, designing teacher training programs, and establishing data governance frameworks that protect student privacy while enabling meaningful instructional personalization. The research will also address policy considerations related to inclusive education standards, equitable access to digital resources, and the integration of sensor-based monitoring with instructional decision-making. The expected conclusion is that AI-assisted personalized PE can reduce participation gaps, elevate physical literacy across diverse learner profiles, and support sustainable inclusive practices in primary and secondary schooling. Recommendations emphasize iterative user-centered design, ongoing educator professional development, and robust evaluation protocols to monitor equity outcomes over time.