Content area
For many beginner programmers, encountering errors in code can be frustrating and disheartening—leading some to questions their belonging in computer science (CS). In these moments, timely debugging help is essential to sustain motivation and foster learning. While students have traditionally turned to peers or teaching assistants for guidance, many now seek debugging support from conversational Large Language Models (LLMs). These chatbots offer promise in providing immediate help, but their ability to generate full-code solutions raises concerns about learning and over-reliance. As these tools become more prevalent, it is important to understand how they can be used to support student's in their debugging and how students seek-help with chatbots.
This dissertation explores how students interact with chatbots in introductory computer science courses (CS1) and opportunities to support debugging. The research is presented in a three-paper format. The first paper examines past debugging interventions before the rise of LLMs, identifying gaps that these tools could potentially address. The second paper presents findings from student interviews about their experiences using a course-integrated chatbot, highlighting how they engage with the debugging assistance throughout the semester and their evolving beliefs about appropriate chatbot use. The third study analyzes naturalistic chat data and survey responses in another CS1 course to investigate how students' goal-orientation and beliefs associate with their help-seeking behaviors. The findings from this dissertation offer insights into designing course chatbots and instructional framing around chatbot use to support students' debugging and learning.
