Um estudo exploratório sobre o uso dos llms como recurso para o aprendizado de fundamentos de programação

Abstract

This study investigates the impact of Large Language Models (LLMs), such as ChatGPT and GitHub Copilot, on programming learning, focusing on the introductory courses of Fundamentals of Programming (FUP) and Introduction to Computer Programming (IPD) at UFC – Campus Quixadá. The research adopts a mixed-methods approach, combining an exploratory study on student approval rates from 2015 to 2024, a survey conducted with students, and interviews with professors teaching these subjects. The results indicate an increasing trend in approval rates since 2021, coinciding with the rise of LLMs, and suggest that these tools are widely used, especially by students with little programming experience. However, their positive impact depends on usage: students who critically engage with LLM-generated responses report greater learning benefits, whereas those who use them passively struggle to consolidate knowledge. Interviews with professors reinforce this perception, highlighting that while LLMs can aid in understanding complex concepts, many students develop excessive reliance on these tools, which may hinder their ability to solve problems independently. Faculty members emphasize the importance of guided and conscious use of LLMs so that these technologies support learning without replacing critical thinking and hands-on programming practice. The study concludes that LLMs can be valuable resources in programming education, provided they are integrated into pedagogical strategies that promote student autonomy and critical thinking.