The use of artificial intelligence (AI) tools and methods is currently one of the main factors defining the directions of the transformation of educational systems, often referred as the Education 4.0 initiative. One of the assumptions of this initiative is the individualization of the learning process (Personalized Learning). It involves defining methods and tools that allow for the adaptation of content and the implementation of the curriculum to the individual predispositions and learning styles of students. This assumption is consistent with another concept in the field of teaching methods, which is referred to as self-regulated learning. This approach is very important in academic teaching, especially in the field of engineering sciences. The process of building new, effective habits and acquiring skills for solving encountered problems in IT subjects is a barrier for a large group of students.

The article proposes a model for conducting laboratory classes, based on the aforementioned combination of AI tools and self-regulated learning. The concept of this implementation has been named as a digital assistant, though a similarly good name is digital twin. The fundamental assumption of the model is to personalize student support in the process of acquiring new competencies while simultaneously reinforcing specific goals defined in an academic course. From the perspective of the model's structure and functionality, this is an extension of the Reinforcement Self-Regulated Learning (RSR Learning) system. This system was described in the article published as part of the previous edition of the ICERI conference (ICERI 2023). A new, key element that is detailed in this article is the introduction of an individual teaching assistant based on Large Language Models (LLM) and data prepared and collected over the past 5 years for the needs of the RSR Learning system. A personalized digital assistant for students allows for tracking and adjusting tasks for each student while maintaining the continuity of the educational process. Preliminary results confirm that the proposed laboratory class model significantly reduces the impact of the aforementioned barriers.

The first part of the article presents the idea and the method of integrating the digital assistant into the existing RSR Learning system. Based on this, the use of this assistant in the process of defining the proposed laboratory class model is presented. Emphasis is placed on identifying aspects of individualization of the LLM model through the appropriate selection of RAG parameters (Retrieval – Augmented – Generation). This is a key issue that, based on the acquired experiences, determine the usefulness of the assistant as well as the complexity of its implementation. The second part of the article contains an analysis of the teaching effects from the last three years (2022 – 2024) based on the example of the cloud computing laboratory. It includes data collected based on the RSR Learning system both without and with the use of a digital assistant. This makes it possible to assess the impact of combining AI tools with the system implementing RSR Learning, as well as the advantages and disadvantages of such an approach. At the same time, the conclusions presented from the analyses may serve as a valuable source of information when planning and implementing similar educational projects.

Keywords: Digital assistant, large language models, self-regulated learning, student motivations, academic laboratories.