Abstract:

This paper presents a conceptual framework for the development of an AI-based recommendation system that selects learning resources to optimize learners’ cognitive load. The goal is to promote sustained engagement and deep learning by aligning the learner’s cognitive effort with the principle of Desirable Difficulty, which postulates that learning is maximized when the level of challenge is neither too low to induce boredom nor too high to provoke overload. The proposed approach shifts the traditional adaptive learning paradigm, from systems that primarily aim to reduce mental effort to one that seeks to calibrate cognitive demand in accordance with each learner’s motivational and affective profile.

The framework provides a multidisciplinary foundation for the assessment of learning competencies through disruptive technologies such as virtual and augmented reality. It combines two theoretical cornerstones: the Cognitive and Affective Model of Immersive Learning (CAMIL), which explains how cognitive processes (attention, memory, decision-making) interact with affective variables (motivation, curiosity, emotional regulation) in immersive learning; and the Evidence-Centered Design (ECD) framework, which ensures that every inference made by the system is supported by valid, systematically collected evidence.

To represent individual differences in affective and motivational readiness, the model leverages four validated self-report instruments: Locus of Control, to assess motivational orientation and perceived autonomy; Emotion Regulation Questionnaire (ERQ), to capture the learner’s strategies for managing emotional states; Curiosity and Exploration Inventory-II (CEI-II), to measure the drive to explore novel or uncertain learning situations; and General Self-Efficacy Scale (GSE), to estimate confidence in one’s ability to achieve learning goals. Together, these measures form a psychological profile that guides the recommendation of learning resources and tasks.

Within the proposed system, these self-reported data are integrated into an adaptive algorithm that dynamically selects digital learning resources, such as VR lessons, AR activities, or multimedia materials, designed to elicit an optimal level of cognitive load. The AI model continuously aligns resource difficulty, interactivity, and feedback mechanisms with the learner’s affective state and evolving self-efficacy. The objective is not to simplify the learning experience, but to induce productive mental effort that fosters metacognition, persistence, and long-term retention. By doing so, the system operationalizes the concept of Desirable Difficulty in an evidence-based and psychologically informed manner.

This work contributes to the theoretical and practical advancement of intelligent educational technologies. It demonstrates how insights from educational psychology, affective computing, and learning analytics can converge to build adaptive systems that balance cognitive demand with emotional regulation. Ultimately, this conceptual model supports the design of next-generation learning environments that are not only adaptive and data-driven but also human-centered, promoting meaningful, resilient, and self-regulated learning experiences.

Keywords:

Adaptive Learning Systems, Cognitive Load Optimization, Education, Recommendation systems.