EDSGR

With generative AI tools like ChatGPT and Copilot increasingly able to produce functional code, programming education faces a new challenge. While these tools can support learning, they also risk encouraging students to bypass core skills such as abstraction, syntax, and problem-solving. Responsible and effective use of AI requires a solid programming foundation, enabling students to critically assess and refine AI-generated output (missing reference).

The EDSGR project proposes developing a dedicated Large Language Model tailored to TU/e programming education. Unlike general-purpose tools, it will align with specific course content and learning goals, offering guidance without directly generating code. Instead, it will prompt reflection, support problem-solving, and help students strengthen their own thinking. The tool will also give educators insight into common learning difficulties by analyzing patterns in student interactions.

EDSGR aims to function as a virtual tutor that supports learning without undermining essential skills. By avoiding automatic code generation, it promotes active engagement and encourages students to reflect on their approach. Students will receive immediate feedback and targeted guidance, helping them navigate complex concepts and develop critical thinking about code.

For instructors, the tool will provide real-time analytics on student progress and identify topics where learners struggle, enabling more focused teaching and improved cohort management. The tool will be piloted in two key Mechanical Engineering courses: Principles of Design and Programming and Scientific Computing for Mechanical Engineering. Evaluation will involve surveys, interviews, and performance comparisons between student groups. Insights from the pilot will guide further development of the tool and teaching strategies.

If successful, EDSGR could be extended to other programming-related courses and departments, offering a scalable way to combine AI support with strong pedagogical oversight.