Interaction Design (IxD) of an Intelligent Tutor for Programming Learning Based on LLM
Oleksiy Levchuk, Carlos Sánchez, Nancy Pacheco, Jesús Favela · Avances en Interacción Humano-Computadora · 2024
The emergent behavior of automatic programming exhibited by Large Language Models (LLMs) has raised uncertainty about the future of programming and its teaching. To better understand this phenomenon, we conducted a field study with programming instructors and students that informed the design of an intelligent tutor to integrate Generative Artificial Intelligence (GAI) into the educational environment. The resulting tool, EVA-Tutor (Virtual Learning Environment), supports the teaching and learning process of programming by establishing bidirectional communication between the student and the LLM through a GPT-4 API and a set of prompts designed to guide and motivate the student with personalized feedback. Rather than solving the problem for the student, the tool helps direct them toward solving it independently. A preliminary evaluation with students and instructors provides evidence of EVA-Tutor's utility and ease of use for problem-solving, knowledge acquisition, and the development of programming skills.