Systems Engineering Approaches for AI-Integrated STEM Learning in Primary Education

Abstract

The operational integration of artificial intelligence within primary STEM education is currently stalled by persistent data silos, high pedagogical latency, and opaque algorithmic governance. To resolve this structural misalignment, this study develops and validates a formal Systems Engineering framework the Architect Model to establish a robust, human-governed data flow within cyber-physical learning environments. Utilizing an engineering-informed systematic literature review protocol, this investigation analyzed 30 high-maturity, peer-reviewed studies (2021–2026) using the Mixed Methods Appraisal Tool (MMAT) and rigorous thematic synthesis. The structural analysis reveals that current platforms frequently induce an efficiency fallacy and severe semantic degradation. These systemic failures are fundamentally driven by proprietary API lock-ins and heuristic limitations that actively disrupt cognitive load management and destabilize instructional stability. To rectify these deficiencies, the finalized Architect Model functions as a definitive system specification, re-engineering the classroom hierarchy by establishing the educator as the Primary Controller. This control architecture mandates a localized metadata registry to guarantee systemic interoperability, enforces explicit operating thresholds to maintain a high signal-to-noise ratio in student telemetry, and deploys modular adaptive feedback controllers to regulate cognitive throughput via closed-loop mechanisms. By subordinating machine autonomy to strict human governance, this framework provides the exact engineering tolerances required to sustain optimal, highly interoperable, and pedagogically secure STEM learning ecosystems.