Cultivating machine learning-related computational thinking in elementary school

There is a growing understanding that Artificial Intelligence education should begin in elementary grades. This necessitates a shift in the perception of what Computational Thinking entails, moving beyond the traditional algorithmic approach to include the data-driven principles fundamental to Machine Learning. This study explores how student activities involving the construction of Predictive Artificial Intelligence systems can be effectively integrated into elementary education, specifically focusing on teaching students to build Machine Learning systems and cultivating Machine Learning-related Computational Thinking. This study used a design-based research methodology with 209 students aged 11-12. The unique year-long, in-school, teacher-led course was based on the constructionist approach. It implements the "Artificial Intelligence for K-12" framework, which aims to systematically incorporate Artificial Intelligence education into the K-12 curriculum, treating it as a unique and essential part of school subjects. This approach focused on actively engaging students in constructing Predictive Artificial Intelligence systems. The results indicate that students demonstrated a notable ability to create Machine Learning systems. They also showed substantial, statistically significant improvement in their understanding of Machine Learning-related Computational Concepts and a high level of ability in using Machine Learning-related Computational Practices. These outcomes, coupled with increased engagement, motivation, and self-efficacy, highlight the value and practicality of introducing Machine Learning concepts early in the educational journey and demonstrate a practical method for implementing the "Artificial Intelligence for K-12" framework in elementary school settings.