Authors:

Ana Maria Hern

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Large Language Models (LLMs), like ChatGPT, offer new learning support in programming education. However, in the absence of clear guidelines, continued usage by students can be risky, as these tools may cause overreliance, limited understanding, or unethical behavior. This study investigates how students enrolled in an introductory programming course interact with ChatGPT when they ask for help. Prior studies have largely focused on outcomes or literature-based analysis; few explore students' behavior and thinking processes from a qualitative perspective on how students actually do, think, and feel while solving tasks with AI support. This research uses real conversations between students and ChatGPT during the course. Through open coding, we labeled expressions of confusion, prompt strategies, and reactions to AI responses; these were refined into broader themes using axial coding. Later, themes were analyzed to understand how students' emotional and cognitive states influenced their prompting strategies and shaped their conversations. Most students relied on the model to solve problems directly, while a smaller group crafted detailed prompts or combined AI support with their own reasoning. Prompts often reflected uncertainty, and reactions varied from being guided to critical questioning, revealing underlying patterns in how students approach problem-solving with AI. These distinct approaches highlight the need for more structured and student-centered guidance on AI use within the curriculum. This study informs the design of instructional strategies, such as scaffolding prompt techniques or fostering reflection, that promote ethical, personalized, and meaningful integration of LLMs into programming education. Keywords: Large Language Models; LLMs; ChatGPT; Programming Education; Student-AI Interaction

Source:

Purdue University / 2025

Topics:

No topics listed

Co-authors:

Ana Maria Hern