Students learning Object-Oriented Programming (OOP) usually find relationships among classes difficult to differentiate. Aggregation, composition, dependency and inheritance are the main relationships used in OOP. Students should master these relationships in order for them to create sound programs. However, when students first start with the OOP paradigm, they have usually previously studied Structured Programming (SP), which introduces concepts that are also used in OOP. This fact makes relationships in OOP confusing for students as they find that the small differences between these relationships can make a huge impact in the final outcome of their program.

Doubts among students usually arise when they are working designing an OOP program or when they are programming it. At those times, students are usually not in contact with a teacher that can solve any misconception or doubts they might have. Intelligent Tutoring Systems (ITSs) have been proposed to help students in those cases. However, these intelligent tutors are still under research and development. Nevertheless, an bounded alternative can be found with the use of Generative Artificial Intelligence (Gen AI).

In this paper we propose the use of Generative AI as a not-specialized tutor for learning relationships in OOP, namely, GenTutor. Two undergraduate groups of “Introduction to Object-Oriented Programming” took part in this study. One was a control group and the other used GenTutor to help them when working on their homework. There were a total of 3 activities that students had to submit during their course. Students in the GenTutor group were directed to ask their questions and to double-check their answers with the GenTutor. Results show that, although differences were not significant, there was a trend of students having a deeper understanding of relationship concepts in OOP. Also, students reported feeling supported by GenTutor as it reassured them the correctness of their answers before submitting their work. Finally, and surprisingly, analysis of students interactions with GenTutor showed that it validated students questions, but it answered incorrectly due to the malformation of the initial query to GenTutor.

Keywords: Generative Artificial Intelligence, Tutoring System, Tutoring, Undergraduate education, Artificial Intelligence.