Large language models in software development learning

This article presents a systematic analysis of the application of large language models in academic software programming education. The findings reveal several thematic domains: virtual tutors for code generation and analysis, feedback and program evaluation, and experimental studies with students, where language models are employed in various ways to support programming courses. Other recurring topics related to the core focus of this research, as identified in the analyzed literature, include computational thinking, ethics in AI-assisted coding activities, and the use of pre-trained models specialized for software programming. Most of the studies reviewed here agree that generative artificial intelligence serves as a significant support tool for achieving programming competencies. However, some research also emphasizes that improper use of intelligent models hinders learning outcomes and fosters unethical behaviors. The thematic domains identified in this review of scientific literature are closely interconnected and essential for understanding the current phenomenon in programming education supported by AI technology. Scientific literature trends indicate that programming education assisted by large language models is an active and growing research area.

Generative grammar · Programming paradigm · Software development · Software development process · Thematic analysis · Software Engineering Research · Software Engineering Techniques and Practices · Teaching and Learning Programming · Software