<![CDATA[General Background: The global educational shift requires movement from surface learning toward complex problem-solving supported by intelligent systems. Specific Background: Adaptive learning grounded in Cognitive Load Theory is increasingly deployed to address diverse working memory capacities through data-driven adjustment. Knowledge Gap: Current implementations remain dominated by behavioral personalization while overlooking internal cognitive dynamics during learning processes. Aims: This study synthesizes empirical evidence to examine mechanisms of cognitive personalization, evaluate its contribution to deep learning indicators, and identify pedagogical constraints through a Systematic Literature Review. Results: Analysis of 13 studies reveals that adaptive systems operate as cognitive architects by dynamically reducing extraneous cognitive load, fostering higher-order thinking skills, improving retention, and stabilizing learning emotions. However, challenges arise from dual-task interface complexity, limited teacher data literacy, infrastructure constraints, and ethical concerns regarding data privacy and algorithmic bias. Novelty: The study advances a Human-in-the-Loop framework integrating AI with fading strategies to prevent analytical dependency while maintaining learner autonomy. Implications: The findings extend Cognitive Load Theory through dynamic load management and highlight the necessity of reconstructing teacher competencies toward AI-Pedagogical Content Knowledge while encouraging future validation across diverse learner contexts. Highlights• Dynamic load reduction mechanisms support sustained conceptual understanding and retention• Intelligent scaffolding increases higher-order reasoning but requires gradual assistance withdrawal• Implementation barriers relate to interface complexity, educator readiness, and system accessibility KeywordsAdaptive Learning; Cognitive Personalization; Deep Learning; Cognitive Load; HOTS]]>
