In the upcoming 2025/2026 school year, we want to to pilot several lessons developed within our project in as many schools as possible across Slovenia, Luxembourg, and Ireland. We have selected four activities and transformed them into engaging challenges. Each challenge is based on a hands-on activity that doesn’t require a computer but connects to a specific area of artificial intelligence. Afterwards, we show students how a computer approaches the same challenge, or how AI-based tools can help solve it.
The challenges are designed for students in the upper grades of primary schools.
No prior knowledge of computer science or AI is required for teachers.
The challenges can be incorporated into different subjects. They naturally fit into mathematics, one aligns with geography, and in the past, teachers have included them in other lessons and activities as well.
Each challenge requires about 1-2 periods. Students do not use computers, except for the first challenge, which requires 4-5 tablets.
Students pick their favorite cartoons and compare their choices with classmates to create a sociogram (“taste map”) that shows who has similar preferences. Based on this, we are able to generate recommendations for cartoons the students might enjoy watching as well.
This activity introduces students to recommendation systems, such as those used by video streaming sites (YouTube), social media platforms (Instagram, TikTok), and online stores.
(Link to the description on the Pumice page)
Students are shown pictures of gnomes, each associated with a known profession: miner, gardener, mason, or tailor. They must identify the underlying rule and predict the professions of new gnomes.
This simple but engaging activity illustrates the essence of artificial intelligence and machine learning: building models. Students actually construct a formal model - a decision tree.
(Link to the description on the Pumice page)
Students sort cards with average temperatures and precipitation from different locations into groups. First, they do this for Europe, then, for a greater challenge, for locations in Asia. This helps them identify regions with similar climates (e.g. dry summers, some winter rainfall) and, if successful, name them according to what they later learn in geography.
The underlying concept is clustering - a method for grouping similar data points. Similar techniques are used across sciences and economics, from biologists studying ecosytems, to marketers analying consumer behavior.
(Link to the description on the Pumice page)
In the final and most exciting challenge, students build a “computer” capable of plying a board game, powered not by electrons but by candies. At first, it performs poorly, but it quickly learns from its mistakes and, after several games, never loses again.
This demonstrates reinforcement learning, where a system learns from experience and feedback. Reinforcement learning is used, for example, to train AI systems like large language models (e.g., ChatGPT) to provide useful responses.
(Link to the description on the Pumice page)
We warmly invite you to join us!