AI and personalized learning: bridging the gap with modern educational goals

TL;DR

The paper analyzes how AI-driven personalized learning aligns with OECD Learning Compass 2030, arguing that current adaptive systems emphasize efficiency and domain knowledge at the expense of learner agency, self-regulated learning, and general competencies. It reviews the benefits of adaptive learning (flexible pacing, immediate feedback, SRL scaffolding) and highlights the mismatch with modern educational goals, including limited domain breadth and reliance on gamification. GenAI is explored as a potential remedy, capable of enriching SRL and collaboration when designed thoughtfully, but risks cognitive offloading without proper safeguards. The authors advocate a hybrid human–AI learning model that foregrounds SRL development, social learning, and teacher facilitation to realize a holistic, lifelong-learning paradigm with equitable benefits.

Abstract

Personalized learning (PL) aspires to provide an alternative to the one-size-fits-all approach in education. Technology-based PL solutions have shown notable effectiveness in enhancing learning performance. However, their alignment with the broader goals of modern education is inconsistent across technologies and research areas. In this paper, we examine the characteristics of AI-driven PL solutions in light of the goals outlined in the OECD Learning Compass 2030. Our analysis indicates a gap between the objectives of modern education and the technological approach to PL. We identify areas where the AI-based PL solutions could embrace essential elements of contemporary education, such as fostering learner's agency, cognitive engagement, and general competencies. While the PL solutions that narrowly focus on domain-specific knowledge acquisition are instrumental in aiding learning processes, the PL envisioned by educational experts extends beyond simple technological tools and requires a holistic change in the educational system. Finally, we explore the potential of generative AI, such as ChatGPT, and propose a hybrid model that blends artificial intelligence with a collaborative, teacher-facilitated approach to personalized learning.

Figures (3)

• Figure 1: a Adapted illustration of the experimental setup of Held and Hein (1963) in which two kittens were reared in darkness from birth and learned to see the world only in a carousel. While kitten P was passively moved around in a casket, kitten A could rotate around on its own. Significantly, only kitten A, exhibiting self-directed movement, developed normal vision. b Illustration of two learners in an open educational field. Learner P is moved by an adaptive learning system (the cross with attached strings) along a prescribed learning path (light gray line) consisting of knowledge bits (darker gray dots) in the fixed knowledge space (circle). Learner A, not bound to the fixed knowledge space, is self-directed and freely explores the educational field, leaving a record of past activities (black dotted line). Like the kittens, only learner A would develop agency and self-regulated learning skills.
• Figure 2: a The prevailing view of personalized learning (PL) as “individualization,” where an adaptive learning system tailors learning content to individual learners to improve their performance. This approach is mainly effective in improving the learning of subject-based knowledge. b The goals of modern education require a collaborative learning environment where learners can share ideas, receive and give feedback, and co-regulate. Figure adapted from (Winter, 2018) with permission from John Wiley and Sons.
• Figure 3: Human instructors are needed to support learners with developing self-regulated learning skills. Scaffolded assistance is required as a function of learner’s age: younger students require more guidance and a social learning environment, while older students require less support. AI assistants are more effective for learners who already exhibit self-regulation skills.