Abstract:

AI-assisted programming tools have become part of how students approach typical software development and database coursework in various computing degrees. However, much of what we know about the use and impact of AI assistance comes from studies of introductory programming courses and general-purpose coding tasks. Considerably less attention has been paid to how these AI tools perform in intermediate- and advanced-level software development courses, where students work with more complex programming language constructs, full-stack frameworks and libraries, and contemporary databases to build applications resembling real-world software. Equally underexplored is how students in such courses perceive and use AI tools in their own learning.

This paper presents an exploratory study conducted across several undergraduate software development and database courses at our institution, covering topics ranging from intermediate-level programming in Java and the use of applied data structures and algorithms, database connectivity and access to common relational databases from Java, to complete enterprise-level database application development using contemporary databases. These courses require students to work with various data structures and collections, and the corresponding algorithms, as well as database integration, first via JDBC and prepared statements, and later with data mappers and object-relational mapping (ORM) tools, all the way through to full-stack database applications based on common frameworks and architectures. These are tasks of a different character than the general coding problems typically used to evaluate AI assistants, and understanding how well current AI tools handle such tasks is important for both instructors designing course exercises and assessments and students deciding how to use these AI tools efficiently in their studies and professional careers.

The first part focuses on an exploratory comparison of selected AI-assisted programming tools across several categories, including general-purpose conversational assistants, IDE-integrated coding agents, and specialized database-oriented AI tools, evaluated against representative tasks from actual course exercises. We primarily examine the correctness, completeness, and explanatory quality of the generated output across tasks of varying complexity. The second part presents our findings from a student survey specifically investigating which AI assistants students use in programming assignments, how frequently, and what they believe the effect of that use has been on their understanding of the course material. We discuss the findings alongside recent research on AI assistance capabilities and learning outcomes in modern software development education, and conclude with practical instructional recommendations to help instructors and students use AI tools in ways that support rather than substitute for genuine learning.

Keywords:

AI-assisted programming, Java, database programming, AI capability comparison, software development education, IDE-integrated AI agents, student perceptions.