Decoding Student Approaches: Navigating Complex Open-Ended Engineering Problems with Large Language Models

The resolution of complex open-ended problems is a core aspect of engineering practice. This study aims to explore students' approaches to using large-language models (LLM) to tackle such problems in engineering. LLMs are highly versatile tools with immense potential in education; however, their effective use to solve open-ended problems requires students to evaluate and analyse information, consider diverse perspectives, and make informed decisions. This raises important questions about students' readiness to critically and responsibly use LLMs for these tasks, as well as the role of educators in fostering critical thinking skills while integrating LLMs into engineering courses. Using a mixed inductive and deductive coding approach to analyse dialogue records with LLMs from nine students and their responses in semi-structured interviews, this study uncovers several misconceptions about LLMs that hinder their effective use. Furthermore, we propose practical strategies for designing problem-solving tasks that enhance students' understanding of LLM capabilities and guide them toward responsible, critical engagement with these tools.