Abstract:

AI tools are ubiquitous in professional programming and are also used by students to solve programming tasks. Instructors must therefore consider how they can meaningfully integrate AI into their teaching without compromising learning outcomes. Early exposure to AI in software development education offers several advantages: It increases relevance to professional practice, boosts productivity, and creates space for core topics such as algorithms, design, and architecture.

Therefore, at our university a programming course was redesigned: This course no longer teaches the pure syntax of a language but focuses on understanding software code, its testing, and good design. The goal is to enable students to critically examine AI-generated code, use it effectively, and adapt or correct it themselves as needed. Students are guided through solving programming tasks with AI support. This exposes them to solutions that, while functional in principle, are not always appropriate or error-free. Students actively learn to read, test, and debug code. Through reflection and creative application, the course aims to demonstrate that AI is a tool, not a replacement for individual creativity. In this way, AI-supported programming are taught in a realistic manner.

The new course was piloted last semester and a survey was conducted at the end to gather feedback for potential continuation. In this paper, the main findings will be presented in details, e.g., the AI usage by the students, their learning progress, positive aspects and challenges from student’s point of view. In summary, the survey presented a predominantly positive picture of the AI-supported course with increased motivation, less frustration, and more flexible learning design. At the same time, it provided clear indications of the need for conscious use and critical review of the generated code.

Also the teacher’s perspective will be described in this paper. The final exam’s results partially contradicted the student’s feedback: Although the students said, that they always critically examined the AI responses, they were often unable to explain their own source code in detail. This underscores the need for targeted skills: Teachers must help learners understand AI-generated code and integrate it into familiar concepts.

A gradual introduction to AI models, typical sources of error, and effective prompt formulation is crucial to prevent students from adopting AI instead of critically reflecting on it. Concrete measures for the teachers include: demonstrating example prompts, incorporating reflection tasks on prompt strategy, and establishing continuous monitoring of AI usage. This allows teachers to provide targeted support for learning processes and translate AI code into understandable, instructive building blocks. These and some further observations made during the course by the teacher will be discussed in this paper.

Keywords:

Software development, AI-generated source code, programming education.