Recommender systems in complexity-informed curriculum design: A framework for enhancing personalized learning

Abstract

This study addresses the persistent limitations of traditional educational recommender systems (ERS), which often fail to model the dynamic, non-linear, and emergent nature of learning. We propose a complexity-informed theoretical framework to reconceptualize the design of these systems, aiming to enhance personalized learning by treating educational technologies as coadaptive ecosystems. To this end, the study addresses four research questions: (1) how the core requirements of personalized learning expose the limitations of static ERS; (2) how principles of complexity theory can redefine their functional architecture; (3) which contextual, behavioral, and pedagogical components most influence their effectiveness; and (4) what algorithmic and design characteristics enable emergent and adaptive behaviors. A structured theoretical review was conducted, involving a systematic search of academic databases from 2008 to 2025 and a two-phase conceptual content analysis. The findings indicate that complexity principles—such as emergence, self-organization, and feedback loops—can fundamentally transform ERS from static filtering tools into dynamic systems that co-evolve with learners. Key characteristics of such systems include adaptive architectures, decentralized processing, and integrated feedback mechanisms. This framework provides a new design logic for creating educational technologies that are more responsive, resilient, and context-aware. Future research must prioritize empirical validation, ensure ethical safeguards such as privacy and transparency, and develop culturally localized models to ensure equitable and effective personalized learning experiences.