Sketch, Predict, Reflect: Two Prototypes of an Interactive Child-Friendly AI Exhibit

• In Event: AERA Poster Session Friday 3:45pm
  In Poster Session: Sketch, Predict, Reflect: Two Prototypes of an Interactive Child-Friendly AI Exhibit

Abstract

This proposal presents an interactive exhibit design to promote AI literacy among children. Intended for a children’s museum, the tangible design facilitates engagement with AI concepts through hands-on drawing and critical interpretation of machine-generated feedback. The design invites children to draw a cat or dog on a sketchpad. An AI model provides a classification, accompanied by confidence indicators and pixel-level spotlights to expose and explain AI’s decision-making. By lowering the threshold for exploration and tacit prompting of follow-up interactions, the exhibit aims to reveal how machine perception differs from human expectation and invites experimentation and reflection on AI’s decision-making processes. We showcase two design iterations and share preliminary insights from children participants in recent pop-up exhibits.

Authors

• Sihan Cheng, Northwestern University

• Lexie Zhao, Northwestern University

• John Chen, Northwestern University

• Michael S. Horn, Northwestern University