Abstract:

The Bebras competition offers a large repository of well-structured tasks that foster computational thinking among students of different age groups. The international Bebras community creates more than two hundred new tasks every year, including both original problems and adapted versions of existing ones with new contexts or difficulty levels. While many tasks can be reused or modified with minor adjustments, some are less suitable for modification.

This paper introduces a taxonomy of Bebras tasks based on their potential for modification. The proposed categorization considers both the contextual features of each task and the nature of the data and their representations. The taxonomy helps identify tasks suitable for remixing or adaptation. Such modifications can be proposed by experts in computer science who carefully evaluate the scope, properties, and pedagogical alignment of the modified task.

Recent advances in generative artificial intelligence enable the use of AI models to support the modification of competition tasks. The paper presents a prototype workflow for AI-assisted task modification, combining language models for content generation with expert validation of the modified tasks. The results of this research support more efficient task reuse not only within the Bebras competition but also in computer science education more broadly.

Keywords:

Bebras competition, task modification, artificial intelligence.