Exploring the Role of Large Language Models as Artificial Tutors

As large language models (LLMs) become increasingly integrated into learning environments, their potential to enhance or hinder the acquisition of computational skills remains debated. This study investigates the role of generative AI (GenAI) tools, particularly Harvard's CS50 Duck, in supporting programming education. Through a mixed-methods approach, we examine students' perceptions, engagement, and practical application of the CS50 Duck within our Computational Thinking course. Our results indicate that while proficient students use GenAI tools to reinforce problem-solving skills, struggeling students may over-rely on them, potentially bypassing critical learning processes. Survey and assignment data suggest that students value the non-judgmental feedback provided by the CS50 Duck, yet express nuanced views on GenAI's role in formal assessments and programming education. We show that analysing chat histories can serve as a qualitative framework for examining the interactions between students and artificial tutors, while simultaneously offering critical insights into students' learning processes and the challenges they encounter. This study also underscores the need for instructional strategies that guide responsible GenAI use and highlights the importance of educator involvement in integrating these tools effectively.