Abstract:

This study examines fostering computational thinking among fifth-grade primary school students (aged 10-11) by introducing artificial intelligence (AI) and machine learning concepts, specifically decision trees for image classification, through unplugged activities. As AI integrates into society and education, early exposure builds digital/AI literacy, algorithmic reasoning, and critical skills without programming or digital tools. The research involved 41 primary school students across three groups: two regular classrooms and one extracurricular workshop group, all of whom participated with parental consent. The intervention employed a combination of task-based assessments and attitude surveys. Initially, students were introduced to foundational AI topics, including machine learning and image classification, with a special focus on how decision trees function as rule-based classification algorithms. Unplugged activities provided the instructional core: students were tasked with constructing their own decision tree to classify images of vehicles, followed by classifying new images according to a predefined tree. These activities deliberately excluded computers, emphasizing hands-on learning and direct engagement with concepts. Supplementary surveys measured students’ attitudes toward AI and their perceptions of decision-tree-based reasoning. Results demonstrate that most students, irrespective of gender, class, or participation group, were able to understand and apply decision tree logic to simple classification problems. Specifically, 61% succeeded in creating their own decision tree and 71% correctly classified images using a provided tree. Attitudinal data revealed high interest in AI and positive reception of decision tree activities, with 4.27 out of 5 students expressing enjoyment in using this approach. No statistically significant differences emerged among groups divided by school context or gender, suggesting a generally broad accessibility of this unplugged AI teaching method. Qualitative analysis revealed most common errors involved using subjective criteria, incomplete hierarchical splits, and an overreliance on everyday knowledge rather than provided data, indicating where further instructional scaffolding is needed. The findings indicate that unplugged activities can demystify AI for young learners and make foundational concepts tangible, even before formal programming skills are in place. Early introduction to AI can thus contribute meaningfully to both digital literacy and analytical thinking development in primary education. The study provides concrete insights for curriculum designers and teaching practitioners aiming to integrate AI concepts into primary curricula through age-appropriate, engaging pedagogy, and underlines the need for continued innovation and research in foundational AI education at all schooling levels.

Keywords:

Artificial Intelligence (AI), Decision Trees, CS Unplugged, Computational Thinking, Primary Education.