The Emerging Computational Thinker

Abstract

As society becomes increasingly shaped by digital technologies, Computational Thinking (CT) has gradually been recognized as an essential 21st century skill and a relevant approach to problem-solving. Consequently, several countries around the world are incorporating CT into elementary and secondary educational curricula (K-12), aiming to provide students with a structured thought process that uses an algorithmic approach for the resolution of problems. However, introducing students to the algorithmic methods of programming has proven to be a complex endeavor that faces challenges on multiple levels. This thesis examines some of these challenges through a series of exploratory case studies that investigate the intersection of evolving computational paradigms and interaction modalities and how these factors can influence how students engage in computational problem-solving. The research is grounded in a compilation of five studies, tracing an empirical trajectory of case studies carried out on authentic educational settings with students. These studies implemented workshop sessions exploring different methods for problem-solving ranging from the use of tangible robotics and visual rule-based programming to activities that utilized computational methods based on trained machine learning models, generative AI and prompt-based programming. By deploying constructionism, an expanded taxonomy based on Shneiderman’s semantic/syntactic model of programming behavior, human-AI interaction perspectives, and established CT frameworks as analytical lenses, this work scrutinizes how different specification modalities and computational paradigms can mediate problem-solving and shape entry barriers for novice learners. The findings suggest that while unassisted visual programming environments can mitigate syntactic friction, the operational demands related to the construction of algorithmic logic remain a major limitation. When shifting to the introduction of Large Language Models (LLMs) for engaging in AI-assisted programming through prompts, the operational demands of the programming process was partly delegated to the AI agent. Nevertheless, the analysis suggests that although natural language interfaces can enable students to bypass rigid coding constraints, the operational demands related to goal identification and formulation, as well as the evaluative process required to manage the construction of a functional algorithmic solution, make human agency essential for successful problem resolution, even under AI-assisted modality. Ultimately, this thesis aims to provide exploratory insights regarding how the traditional notions and scope of CT education must be re-defined and expanded. This re-definition is necessary to enable educators to design and operationalize CT interventions that equip students with the new skills required to navigate the wide range of possibilities that emerging computational technologies can offer to engage students in problemsolving within contemporary K-12 education.