Large Language Models (LLMs), a class of Artificial intelligence (AI) tools, are reshaping programming education, allowing novices to take on tasks that go beyond traditional CS1-style problems. While these tools lower barriers, little is known about the factors that shape students’ successful programming or learning in such contexts.

My research investigates how novices use LLMs to complete programming tasks that demand both domain-specific and effective prompting. Building on prior work, I examine how programming background, amiliarity with libraries libraries and their associated application programming interface (API), and prompting behaviors influence students’ coding performance and learning.

As a second-year PhD student, my work is still in development. I expect to contribute empirical insights into AI-assisted programming, characterize common strategies and challenges, and inform the design of supports that promote not only effective code generation but also meaningful learning for novices and more advanced learners alike.