Related Experiment Video
Updated: Aug 6, 2026
03:14
Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
Published on: December 6, 2024
The adaptive learning taxonomy for responsible large language model integration in higher education
Simon Baradziej1, Adrianna Kochanska2
1Department of Technology and Safety, Faculty of Science and Technology, University of Tromsø (UiT) - The Arctic University of Norway, Postboks 6050, 9037 Tromsø, Norway.
Discover Artificial Intelligence
|July 26, 2026
Summary
Large Language Models (LLMs) in higher education show performance gains but risk reduced metacognition. A new framework, ALT-ED, guides responsible LLM adoption by focusing on metacognitive development and data minimization.
Area of Science:
- Educational Technology
- Artificial Intelligence in Education
- Learning Sciences
Background:
- Large Language Models (LLMs) are increasingly adopted in higher education, but institutions lack frameworks for responsible integration.
- While LLMs enhance academic performance, a 'cognitive paradox' emerges, with potential declines in metacognitive accuracy and self-regulation.
- Existing frameworks do not comprehensively address LLM adoption across adaptation design, data governance, pedagogical strategies, and learner agency.
Purpose of the Study:
- To develop a theoretically grounded framework for the responsible adoption of Large Language Models (LLMs) in higher education.
- To address the 'cognitive paradox' of LLM use, balancing performance gains with metacognitive development.
- To structure institutional decision-making for LLM integration across key operational dimensions.
Main Methods:
- A narrative synthesis of 35 empirical studies (Jan 2024-Feb 2025) informed the framework development.
- The framework is grounded in Connectivism, Distributed Cognition, Cognitive Load Theory, and Self-Regulated Learning.
- The Adaptive Learning Taxonomy for Educational Decision-Making (ALT-ED) was developed, structuring decisions across Adaptation Trigger, Data Granularity, Pedagogical Locus, and Agency.
Main Results:
- The 'cognitive paradox' necessitates shifting the pedagogical locus towards metacognitive development.
- High-resolution data collection can amplify algorithmic bias, supporting progressive data minimization.
- Transparent, auditable design frameworks are crucial for bridging the policy-practice gap, rather than prohibition.
Conclusions:
- Prioritize a metacognitive pedagogical locus to foster "learning to learn" skills.
- Employ progressive data granularity to balance personalization with student privacy.
- Default to learner-negotiated agency to safeguard against algorithmic determinism.