Towards a process-oriented assessment of computational thinking: behavioural data and machine learning approach
Vaida Masiulionytė-Dagienė, Tatjana Jevsikova · Interactive Learning Environments · 2025
Computational thinking (CT) is increasingly recognised as a vital skill for the twenty-first century. However, the lack of reliable assessment tools continues to create significant challenges for assessing students' CT skills effectively. This study uses behavioural data from the interactive Bebras Challenge task to assess CT skills, offering a deep analysis of the problem-solving process beyond final binary outcomes. The dataset of solutions (N = 336) from an interactive CT task, focusing on recorded behavioural data, such as object interactions, decision sequences, colour choices and time taken to complete the task, was analysed. Machine learning algorithms were applied to cluster the solutions, identifying distinct patterns that reflected different problem-solving strategies. The analysis went beyond binary classification, revealing variations in solution approaches within successful and unsuccessful attempts, allowing for a more detailed assessment of CT skills. The findings from this study suggest a scoring system based on identified solution clusters as a more refined assessment of CT compared to existing binary methods. This underscores the potential of behavioural data to improve CT assessment.