A Common Framework for Automatically Assessing Computational Thinking Skills

Approaching problems in a logical and structured way is currently considered a key skill for finding appropriate solutions to complex problems. Particularly in Computer Science degrees, introductory programming courses aim to teach programming methods and techniques, as well as to develop a way of thinking that enables students to generate good solutions to diverse problems. The term computational thinking (CT) encompasses a set of mental processes for solving problems using skills such as abstraction, algorithmic thinking, and problem decomposition, among others. While CT research has largely focused on primary and secondary education, often using tools like Scratch or Bebras tasks, studies targeting university-level students remain limited. Training and assessment of CT in higher education, therefore, present an open area of interest for research. This paper proposes a novel approach for automatically analyzing students' source code to evaluate their development of CT skills. We introduce a framework grounded in software engineering metrics and validate it through an initial case study. The findings underscore the need for a second study involving intentionally designed exercises to assess CT competencies more effectively.