Overview: Youth can benefit from more opportunity to experiment with mechanics of training data in AI to understand its workings and ethical impacts. Transformational and educational games are an under-explored, engaging way to help youth explore complex and serious topics with AI. This is an un-plugged multi-player card game that you can play in the classroom or at home! The theme of dogs is fun for all ages.
Game Description: We designed the game with learning science-backed design principles, as well as multiplayer aspects to enable benefits from peer-to-peer and intergenerational learning. Data Detectives supports understanding the basic mechanics of training data, and then gets players to think about how different training data can impact AI outcomes and broader societal impacts.
Notes:
Below are two versions of the game. One with real training data pictures with dogs and another with fictionalized humanoid animal characters. We found that the humanoid animal characters help with better transfer of understanding socio-ethical concepts.
Audience: Middle school-aged youth or older and peers / family.
Research Publications:
Data Detectives is designed to support social learning and make learning the mechanical aspects of AI less intimidating
Players embody two roles the 'Training Data Master' and the 'Algorithm,' where the training data is sorted with a secret rule that the Algorithm must guess
VERSION 1 Slide deck with an introduction to training data and Data Detective game instructions.
VERSION 1 Cards to print and cut out. These images are from real, public machine learning data sets.
VERSION 2 (improved game elements) Slide deck with an introduction to training data and Data Detective game instructions.
VERSION 2 Card deck made with characters from animal crossing.