AI-Powered Adaptive Learning Systems: Theory, Applications, and Ethical-Legal Perspectives in Early Childhood, School, and Higher Education
Author: Boris Mihaylov
Affiliation: Creativity Ltd.
Abstract
AI-supported Adaptive Learning Systems (ALS) are fundamentally reshaping the design, delivery, and assessment of educational processes. This paper synthesises the theoretical foundations, technological mechanisms, and empirical evidence for the effectiveness of ALS across three educational levels—kindergarten, school, and university. Key adaptation models, pedagogical strategies, ethical and legal frameworks (incl. Regulation (EU) 2024/1689—“AI Act”), and best practices for sustainable implementation are presented. The analysis shows that ALS contribute to personalised learning, increased engagement, and higher academic outcomes, provided that the principles of transparency, fairness, and data protection are observed.
Keywords: adaptive learning systems, artificial intelligence, personalised learning, AI Act, ethics in education
1 Introduction
The paradigm of the “one-size-fits-all curriculum” is gradually giving way to personalised learning trajectories driven by data and algorithms. Adaptive Learning Systems, defined as digital platforms that dynamically adjust content, sequencing, and scaffolding to the learner’s individual profile [1], have become a strategic tool for achieving inclusive and effective education. Numerous empirical studies report heightened learning motivation [2], improved conceptual knowledge acquisition [3], and reduced inequities [4].
2 Theoretical Foundations of Adaptive Learning Systems
2.1 Definition and Core Characteristics
ALS are built on three interconnected models—the learner model, domain model, and pedagogical (instructional) model. The decision-making component (“adaptive engine”) combines the data from the three models and determines an appropriate pedagogical intervention in real time [1].
2.2 Historical Overview
From the first “intelligent tutoring systems” of the 1970s to contemporary cloud platforms based on deep learning, the development of ALS has passed through three key stages: rule-based systems, statistical/IRT-based systems, and systems powered by Large Language Models (LLM) [5].
2.3 Contemporary Adaptation Algorithms
- Knowledge Tracing (KT) – estimates the probability that a learner has mastered a skill at any moment; variants: BKT, DKT, GKT [6].
- Item Response Theory (IRT) – calibrates task difficulty and learner ability.
- Reinforcement learning – optimises pathways through learning content.
- Deep neural networks – process multimodal data (e.g., video, speech) and create rich learner embeddings [7].
3 Technological Foundations and Architectures
A typical architecture includes: (1) a data-collection layer (LMS, IoT sensors, apps); (2) an analytics layer for modelling; (3) an adaptive recommendation layer; and (4) a user interface. Open standards (xAPI, LTI 1.3) provide interoperability among systems.
4 Applications by Educational Level
4.1 Early Childhood Education
Adaptation is realised through game-based environments, adaptive stories, and sensor-equipped devices that regulate task complexity in line with children’s attention and motor skills (e.g., Osmo Little Genius). A pilot study in Australia reports 20 % higher cognitive engagement when such platforms are used [12].
4.2 School (K–12)
Widespread solutions include Khan Academy MAP Accelerator, DreamBox, ALEKS, and government chatbots (e.g., EdChat, South Australia) [12]. A year-long quasi-experimental study with 20 000 students shows 47 % higher-quality responses and significantly increased motivation [12].
4.3 Higher Education
University platforms (IntelliPath, Smart Sparrow, AIIA framework [8]) offer micro-adaptive routes combining video, simulations, and LLM-supported tutoring. A comparative analysis at a Chinese university shows that students working with Yixue Squirrel AI achieve higher results than control groups taught by expert instructors [7].
5 Value and Pedagogical Benefits
- Personalised pace – reduces cognitive overload and maintains an optimal “zone of proximal development”.
- Precise progress data – teachers receive dashboards with success-probability forecasts and intervention recommendations.
- Reduced administrative burden – automated assessment and feedback generation save up to 5 hours of teacher work per week [12].
6 Ethical and Legal Aspects
6.1 Privacy and Data Protection
ALS fall into the “high-risk” category under the AI Act (Regulation EU 2024/1689) and are subject to strict requirements for data governance, accountability, and human oversight [9].
6.2 Bias and Fairness
Algorithmic models can reproduce social and cultural biases. Regular testing with demographically stratified datasets and Explainable AI mechanisms is recommended [10].
6.3 The Teacher’s Role
The teacher is positioned as the “human-in-the-loop”, validating automatic recommendations and providing socio-emotional support.
7 Best Practices for Implementation
- Pilot introduction with clear KPIs
- Integration with existing LMS via LTI/xAPI
- AI literacy upskilling for staff (Art. 29 AI Act)
- Ethics boards and protocols for algorithmic audit
8 Challenges and Future Directions
- Interoperability among different adaptive platforms
- Multimodal learning enriched with XR and biometrics
- Personal agent avatars generated by Large Language Models that provide cognitive and metacognitive coaching
- Regulatory dynamics – full enforcement of the AI Act is expected by 2027, requiring new compliance standards [11]
9 Conclusion
Adaptive Learning Systems demonstrate significant potential for transforming education by delivering evidence-based benefits to learners and educators. However, their successful and ethical integration requires a holistic approach combining technological innovation, pedagogical design, and regulatory compliance.
References
- Artificial Intelligence-Enabled Adaptive Learning Platforms: A Review. ScienceDirect, 2025.
- AI-Driven Adaptive Learning for Sustainable Educational Transformation. Wiley, 2024.
- Personalised Adaptive Learning in Higher Education: A Scoping Review. ScienceDirect, 2024.
- The Role of AI-Based Adaptive Learning Systems in Digital Education. ResearchGate, 2024.
- U.S. Department of Education. Artificial Intelligence and the Future of Teaching and Learning, 2023.
- AI in Education: Adaptive Learning Systems. Nano-NTP Journal, 2024.
- Cui, W. et al. Performance Comparison of an AI-Based Adaptive Learning System in China. arXiv preprint, 2019.
- Sajja, R. et al. Artificial Intelligence-Enabled Intelligent Assistant for Personalised and Adaptive Learning in Higher Education. arXiv preprint, 2023.
- European Commission. AI Act – Regulatory Framework on AI, 2024.
- FeedbackFruits Blog. From Regulation to Innovation: What the EU AI Act Means for EdTech, 2025.
- Babl.ai. Navigating the New Frontier: How the EU AI Act Will Impact the Education and Training Industry, 2024.
- AdelaideNow. How AI Is Helping Teachers Spend Less Time on Non-Teaching Tasks, 2025.
