Students’ Perceptions of an Intelligent Tutor in Secondary Education

The increasing diffusion of Large Language Models (LLMs) is opening new perspectives for their use in educational contexts, particularly as interactive learning support tools. This paper presents results from a study involving the use of an intelligent tutor based on an LLM in two third-year classes in upper secondary school in Italy. The tutor was intended as a support tool for learning, designed to guide students by means of a Socratic-style interaction. In a first phase, a class received a training on Prompt Engineering (PE) and then used the tutor on the deepening learning activities about a topic, while the other class met the same activities without the tutor. In the second phase, also the other class received the PE training and both classes used the tutor on a second subject. We focus on the analysis of the final satisfaction questionnaire, aiming to see how students perceived the use of the tutor, across six perceptual dimensions. Overall, the results show a positive and fairly consistent perception of the tutor’s support, with perceived usefulness, instructional effectiveness, and conscious use of the tool showing slightly higher values, and lower values for engagement and metacognition. These findings suggest a good level of acceptance of the tutor among students, alongside some room for improvement, particularly in terms of engagement. The paper concludes with reflections on the integration of LLM-based tutors into everyday teaching practice, highlighting both opportunities and challenges observed during the study.