January 1, 2025

Using image classification to teach machine learning to gifted elementary school students

Key Points

Students achieved model accuracy from 84% to 96%, showing effective learning through hands-on activities.
Engagement in data collection and model training fostered critical reflection on ethical use and challenges.
Project-based learning utilized Google's Teachable Machine, encouraging collaboration without writing code.
Findings highlight the importance of integrating no-code platforms for foundational digital literacy skills.

Abstract

Artificial intelligence and machine learning are becoming increasingly integral to modern life, yet these concepts remain predominantly absent from elementary education. This study presents a project-based learning activity in which talented students enrolled in the Centre for Talented Youth Kragujevac trained their image classification models using Google's Teachable Machine, a no-code Web tool. During five sessions, students (aged 13-14) engaged in data collection, model training, testing, and reflection without writing a single line of code. The activity aimed to demystify artificial intelligence by making it hands-on, collaborative, and accessible. Students demonstrated strong engagement and produced models with accuracy ranging from 84% to 96% while reflecting critically on challenges such as dataset bias, overfitting, and the ethical use of artificial intelligence. The findings suggest that integrating creative, no-code platforms can foster foundational digital literacy, promote ethical awareness, and spark student interest in emerging technologies.

Mark Helpful

Bookmark

Relay