Codify

An Intelligent Socratic Tutoring System for Programming Education

Abstract

Programming education poses significant challenges for many students due to varying priorities. Traditional classroom instruction often lacks the scalability required to provide personalized support. This paper introduces AI Tutor, an intelligent tutoring system designed to enhance programming education through adaptive, conversational learning. Leveraging large language models (LLMs), competency tracking, and adaptive assessment, the system guides students using a Socratic teaching methodology that promotes discovery-based learning over direct answer generation.

AI Tutor, a comprehensive platform, incorporates several key components. These include conversational tutoring, automated practice generation, competency modeling, code analysis, and gamified engagement mechanisms. The platform dynamically adapts to student performance by monitoring their topic-level competency scores. This allows it to adjust the difficulty of questions and the instructional scaffolding accordingly. Students interact with the tutor through a chat-based interface. The system analyzes their responses, updates mastery estimates, and generates targeted feedback.