Interacting By Drawing: Introducing Machine Learning Ideas to Children at a K-9 Science Fair

Abstract

In the 2010s machine learning (ML) became a key driver of quickly growing number of apps and services. How to teach ML and other data-driven approaches in K-12 education has become a focus of intensive research efforts, with many recent advances in technology, pedagogy, and classroom integration. What to teach learners, at what age, and how, are some of the open questions being explored. This study explored children's interactions with a simple image classification tool, which used only two features to classify images. The results offer a proof-of-concept of how to teach 1) the principles of the ML workflow, and 2) some central ML insights, including image recognition, supervised learning, training data, model, feature, classifying, and accuracy. The results recognize how learning the principles of technology facilitates a shift in the locus of explanation from what oneself does to what the computer does. The results provide examples of how to support children's developing data agency.