AI-Integrated Learning Management System for Middle School: A Longitudinal Study of Learning Outcomes Through High

Synthesis: This paper proposes an AI-integrated LMS designed specifically for middle school instruction, addressing the gap between current LMS platforms (which function as workflow tools) and the need for real-time instructional support during the critical middle-school learning window. Key features include policy-gated AI assistance (bounded AI to avoid Over-Reliance), formative feedback and hinting during practice, spaced review and adaptive practice recommendations, and teacher dashboards for misconception patterns. The system is designed with privacy-first architecture including data minimization, age-appropriate response constraints, auditable logs, and role-based access control. The proposed longitudinal study tracks students from middle school through high school into post-high school pathways, linking fine-grained learning traces to institutional outcomes. This design study is complementary to Intelligent Tutoring systems research and Personalized Learning implementations in K-12 settings.}, The emphasis on bounded AI support — rather than open-ended tutoring — distinguishes this approach from systems like Khan Academy's Khanmigo and reflects pedagogical concerns about Scaffolding versus answer-giving in K-12 AI tools.

What this means for practice

• Instructors. Policy-gate AI assistance by activity type: keep bounded, stepwise hints available in practice mode and switch AI help off on graded items, so that gains show up on work where AI was never available.
• Keep assistance in hint form. The design deliberately supplies small next-step guidance and short curriculum-aligned explanations rather than answers, to hold off the over-reliance that open-ended tutoring invites.
• Use the teacher dashboard to act on misconception clusters and sustained struggle while students are still practicing, instead of discovering the gap at the end-of-unit assessment.
• Drive spaced review and adaptive practice from each learner's mastery history rather than from the pace of the class.
• Administrators. Treat privacy as an architecture requirement before any pilot with minors: data minimization, role-based access control, age-appropriate response constraints, and auditable logs of every AI interaction.

Limitations

• No empirical results exist yet. This is a design proposal, and the authors state plainly that the "results" are the expected outputs of the proposed LMS and the artifacts it is designed to generate — there is no classroom deployment, no student outcome data, and no participants.
• The longitudinal evaluation itself is a plan: the authors specify a cluster-randomized or stepped-wedge rollout as the preferred design, so every claim about effects persisting through high school and beyond is untested.
• The threats the design must survive are named rather than measured — implementation variability, novelty effects, fadeout across educational transitions, measurement drift as assessments and curricula change, and attrition and missing data over a multi-year horizon.
• Coverage is narrow by the authors' own account; they call for broadening the system beyond its initial subject areas and activity formats to writing, project-based learning, and collaborative tasks.

Citation

Etchie, M. P., & Olutosin, T. (2026). AI-Integrated Learning Management System for Middle School: A Longitudinal Study of Learning Outcomes Through High School and Beyond.