Dynamic Analysis of Scratch Projects to Infer Computational Thinking Abilities

Abstract

The role of visual block-based programming languages has become prominent in children's computer science education in many schools across the globe, allowing children to concentrate on creating programs by eliminating syntactical program errors. Consequently, the necessity for auto-assessment systems has become apparent as the evaluating learner' projects required manual labour, which was placed on instructors' shoulders. Thus, numerous auto-assessment systems are built to assist instructors in evaluating students' computational thinking skills to cope with this increasing demand. Inspired by the existing literature review on this topic, we envision that behavioral similarity between Scratch programs and their code coverage could be used to infer the Computational Thinking skills of learners. Therefore, we built a web-based tool called DSEScratch that calculates three metrics of behavioral similarity and code coverage by employing dynamic symbolic execution. We anticipate that our system could complement existing Scratch analysis tools to gain deeper insights into learners' Computational Thinking skills.