This article describes the design science research process involved in developing a new tool for teaching K-12 children about the complete machine learning workflow, specifically with image classifiers. The tool was designed alongside 36 6-hour classroom workshops in 12 schools with 213 4th and 7th grade children, for which research permission was obtained from guardians and the municipality. Additionally, a usability study provides further design insights along with a co-design process involving teachers. The goal is to address the limitations with existing tools that have been critiqued in literature by first detailing the design principles that must guide such tool development.
