Presented at Mensch und Computer 2025

Abstract: Large Language Model (LLM) coding assistants have become ubiquitous in software development environments, and features such as auto-completion and contextual code suggestions are now easily accessible to software engineering students. While these tools benefit productivity improvements for experienced developers, their presence in programming education potentially creates major challenges for foundational skill development. As these tools enable students to effortlessly generate functional code without comprehension of underlying logic or software architecture implications, this trend may preclude novices from developing the algorithmic thinking and problem-solving skills essential for complex software engineering tasks. This position paper discusses potential conflicts between current implementations of LLM assistants and pedagogical goals in software engineering. As a human-computer interaction challenge, how might we redesign these tools to preserve the cognitive engagement necessary for learning while retaining their benefits? Rather than advocating for access restrictions, we argue that the HCI community should lead in developing interaction patterns that encourage reflection and ensure that AI augmentation enhances rather than undermines the educational mission of software engineering programs.