Abstract:

This paper describes the design and early evaluation of a generative artificial intelligence tutor to support learning in an introductory undergraduate data structures course. The project addresses a frequent challenge in programming-related subjects: students need timely and personalised support outside class time, yet instructors face increasing difficulty in providing scalable feedback without reinforcing dependency or superficial trial-and-error strategies.

The proposed tutor was designed as a guided support tool rather than an answer generator. Its pedagogical role was to provide hints, scaffold problem understanding, prompt students to explain their reasoning, and recommend next steps according to the type of difficulty detected. The design emphasised self-regulated learning, formative feedback, responsible AI use, and alignment with course learning outcomes and assessment requirements.

The study reports the first implementation of the tutor in a real teaching context and analyses both educational value and practical limitations. Evidence is collected through system interaction logs, student surveys, selected anonymised conversations, teacher observations, and indicators related to use patterns, perceived usefulness, common misconceptions, and instructional workload. Special attention is given to the tensions between personalisation, academic integrity, overreliance on AI, and the need for human oversight.

Preliminary findings suggest that the tutor can improve access to feedback and help students persist in problem-solving tasks, especially in early stages of learning, while also generating valuable information about recurring conceptual bottlenecks. At the same time, the paper discusses limitations related to prompt quality, hallucinations, uneven student use, and the need to frame AI support within clear pedagogical and ethical boundaries. The experience contributes a realistic and transferable case of AI-supported tutoring in higher education.

Keywords:

Generative AI, Intelligent tutoring systems, Self-regulated learning, Programming education.