Research Summary: Understanding Student Use of Large Language Models Across Computer Science Subfields

A research study involving 211 undergraduate computer science students examined their use of large language models (LLMs) across various subfields. The study, conducted in a problem-solving course designed to foster responsible LLM use, analyzed prompt counts and students' conceptualization of LLM roles based on post-assignment reflection data from seven instructional modules. The findings aim to inform the design of subfield-aware instruction as LLMs integrate further into computing education.

Why it matters

This research provides insights into the evolving landscape of AI tool adoption within educational settings, specifically for technical disciplines. Understanding how students utilize large language models across different specializations is crucial for developing future-ready curricula and ensuring effective skill development for the workforce.

Key insights

• The research investigates undergraduate students' LLM usage across different computer science subfields.
• The study involved 211 undergraduate students participating in a problem-solving course.
• The course was intentionally structured to support responsible and effective LLM use through instruction and reflection.
• Data was collected from post-assignment reflections across seven instructional modules.
• The analysis focused on prompt counts and students' conceptualization of LLM roles.
• The goal is to understand how LLM use varies across technical and pedagogical contexts.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2610.01158