How and Why Novices Use LLM-Based Tools to Solve CS1 Coding Tasks

Abstract

As Large Language Models (LLMs) gain popularity in programming education, understanding how novice programmers use these tools and their motivations is essential. Although many studies have explored AI tools like ChatGPT in computing education, few focus specifically on why students turn to these tools. Through 16 lab sessions incorporating participant observation and a questionnaire, we investigate factors driving novice programming students to use AI tools while completing CS1 coding assignments and examine how students engage with these tools.

We provide findings from a thematic analysis examining (a) what learners ask from AI tools; (b) characteristics of learner-written prompts in terms of clarity and crafting patterns; and (c) how learners use AI-generated responses for learning, modifications, and verification. Our analysis identifies key motivations driving novice programmers to use AI tools: timesaving and task management, ease of access, aided learning, and cognitive miserliness. Furthermore, our analysis reveals diverse engagement patterns, including a novel "AI Conceptual-Hybrid" approach, where students use AI to clarify concepts and task requirements but write code independently. We offer insights on how inexperienced learners utilize AI tools, emphasizing self-regulation, indications of overreliance, and opportunities for improving AI-assisted learning tools in computing education.