This study presents a comprehensive examination of how generative artificial intelligence (GenAI), particularly large language models (LLMs), can be leveraged to enhance personalized learning (PL) in educational settings. The study systematically explores the theoretical foundations, technical mechanisms, application services, and implementation challenges of GenAI-PL, while proposing actionable pathways for its effective integration.
At its core, the research identifies GenAI as a transformative force in reshaping personalized learning by enabling dynamic content generation, adaptive feedback, and immersive learning experiences. The study highlights three key characteristics of GenAI-PL: (1) the ability to generate highly customized learning content tailored to individual learners’ needs, (2) the provision of contextualized and immersive learning experiences, and (3) the support for real-time adaptive assessments. These capabilities are grounded in interdisciplinary theories, including humanistic theory, constructivist learning theory, multiple intelligences theory, embodied cognition, and distributed cognition, which collectively inform the design and implementation of GenAI-PL systems.
The study outlines two primary technical solutions for GenAI-PL: edge-based GenAI solutions and pedagogical theory-guided approaches. Edge-based solutions focus on lightweight, locally deployed models that prioritize privacy and incremental learning, while theory-guided approaches emphasize prompt engineering and knowledge graph integration to enhance reasoning and personalization. The study also details four key application services: precise data mining and performance evaluation, cognitive/non-cognitive diagnosis and content generation, interactive problem-solving with real-time feedback, and enhanced learning experiences fostering creativity.
Despite its potential, the study identifies significant challenges in current GenAI-PL implementations. These include limitations in understanding individual static and dynamic differences among learners, insufficient adaptation to higher-order literacy goals, and gaps in ethical and safety regulations. The study particularly emphasizes the risks of algorithmic bias, over-reliance on technology, and the lack of mechanisms for fostering social interaction and critical thinking.
To address these challenges, the authors propose a six-pathway implementation framework: (1) innovating interdisciplinary learning theories, (2) developing reliable large-model technologies, (3) refining agent-based foundational services, (4) enhancing semantic-aligned support for core competencies, (5) optimizing long-term evidence-based impact analysis, and (6) establishing robust safety and ethical guidelines. This framework aims to balance technological advancement with pedagogical integrity, ensuring that GenAI-PL systems are both effective and ethically sound.
In conclusion, the study positions GenAI as a powerful enabler of personalized learning but underscores the need for careful integration that prioritizes learner-centered design, ethical considerations, and continuous pedagogical innovation. By bridging theoretical insights with practical applications, the paper provides a roadmap for educators, technologists, and policymakers to harness GenAI's potential while mitigating its risks, ultimately fostering a more adaptive, inclusive, and effective educational ecosystem.
DOI:10.1007/s44366-025-0056-9