Developing Intelligent Tutors with Artificial Intelligence for Digital Education

This paper explores the applicability and effectiveness of using large language models (LLMs), specifically the LLaMA2:13b model, in the context of digital education, with the goal of supporting individualized learning and improving access to complex content. In an increasingly digitized educational landscape, integrating AI-powered functionalities has become a strategic element in modernizing teaching and learning processes. The practical implementation was carried out within the Moodle 5.0 platform, through the integration of custom AI actions such as: quiz question generation, text summarization, and content explanation. The model was tested across two university-level courses with distinct profiles: Advanced Multimedia Technologies (with an applied focus) and Data Protection and Open Licenses (with a theoretical and legal focus). For each course, the functionalities were applied to Page and Book resources, evaluating the accuracy, relevance, and clarity of the generated outputs. Qualitative evaluation showed that the LLaMA2:13b model is capable of producing coherent and contextually appropriate content, offering real support in understanding course materials and preparing for assessments. Furthermore, the AI integration contributed to personalized learning and increased educational efficiency. Some limitations were also observed, such as the occasional appearance of the "Something went wrong" error, likely caused by server resource constraints or temporary service instability. Nevertheless, the conclusions of this study highlight the significant potential of AI models integrated into educational platforms like Moodle, and open new directions for future research in the field of AI-assisted learning.