Learning to Program with LLMs: How University Students Use LLMs in an Introductory Programming Course

• In Event: Student Experiences with AI-Assisted Learning

Abstract

This qualitative study explores how large language models (LLMs) supported student learning in an introductory programming learning context. Based on interviews with university students and inductive content analysis, we identified two major areas of LLM use: programming knowledge development and coding practice. Students used LLMs to support conceptual understanding, develop programming templates, plan projects, and implement solutions. Notably, we observed advanced, learning-centered uses—such as template development—that point to opportunities to better align LLM support with students’ higher-order learning processes through instructional design. Students also actively exercised agency in addressing challenges like over-reliance and hallucinations. These findings challenge binary views of LLMs as inherently helpful or harmful and underscore the importance of fostering reflective, strategic use in programming education.

Authors

• Yuwei Liang, University of Pennsylvania

• Shiyang Zhang, University of Pennsylvania

• Bodong Chen, University of Pennsylvania