Research

In introductory programming courses it is essential for students to work on exercises independently. However, especially in blended learning environments there is not enough individual feedback while working on exercises, leading to frustration from time-consuming use of wrong strategies. To counteract this we are testing as a part of the project OTH-Brain a concept linking interactive programming exercises with a tutorial AI assistent in a blended programming course for international students.

The key ideas of the concepts are:

• Interactive programming exercises with automatic feedback 
  The exercises are developed in CodeRunner, a system that enables automatic (formative/summative) assessment and immediate feedback on the current solution attempt.
• Continuously trained, module specific AI assistent 
  An AI assistent that is configured for the course is being integrated into the learnig management system and supports students with exercise completion. The underlying model is fine-tuned with regular use.
• Analysis of conversation data for continuous improvement of the exercises 
  The anonymised enquiries to the assistant are analysed on a task specific basis. This way exercise tasks can be specifically improved.
• Accompanying project seminar "BrainLab – AI4Peers" 
  Students of the study programme "KI" (Künstliche Intelligenz - Artifical Intelligence) actively participate in designing and developing the AI tutor in a project seminar.

The project is funded by Freiraum-Förderung of Stiftung Innovation in der Hochschullehre. The project started 1 April 2026 and runs until 31 March 2028.