The future of personalized learning with AI

Abstract

Personalized learning has the potential to enhance learning outcomes but faces challenges such as problems measuring learning variables, the lack of theoretical guidance for interventions based on these variables, and ethical considerations. The availability of new AI tools, including generative AI based on large language models (LLMs), has resulted in claims that the problem of personalization has been solved. The goals of this paper are to provide a shared, systematic language for describing adaptivity, adaptability and personalization of learning environments; to broaden how we conceptualize personalization; to define the adaptation cycle; and to discuss which aspects of the cycle can be supported by AI. Also discussed are levels of AI integration, and a proposed architecture for AI-based personalized learning environments.

Education relevance

The personalization of learning, i.e., the accommodation of every person's individual needs during the learning process, has been a goal for educators and learning environment designers alike. However, achieving this goal has been a challenge as key aspects of the personalization cycle face challenges, such as the measurement of learning variables and the lack of theoretical guidance for interventions based on these variables. We also need to consider ethical considerations related to unintended consequences of personalization. In this paper, we aim to deepen the discourse about personalized learning by defining key terms, showing how to broaden current conceptualizations of personalization to include affect, motivation, and socio-cultural variables, and providing terminology for the adaptivity cycle that allows for a more conceptual clarity. We then discuss how AI can support personalization of learning, with a focus on applications beyond chatbots. We introduce levels of AI integration in personalization and discuss an architecture that may enable personalization of learning through generative AI.