Rapid Testing, Duck Lips, and Tilted Cameras: Youth Everyday Algorithm Auditing Practices with Generative AI Filters

TL;DR

This study explores how high school youth actively engage in algorithm auditing practices with generative AI filters on TikTok, revealing their sophisticated testing methods and potential for integrating everyday and formal auditing approaches.

Contribution

The paper uncovers youth's detailed algorithm auditing practices with generative AI filters, highlighting their capacity for systematic testing and informing design strategies for educational tools.

Findings

Youth conduct extensive testing of AI filters using camera variations.

Participants identify limitations and biases in generative AI filters.

Young users demonstrate sophisticated understanding of AI behaviors.

Abstract

Today's youth have extensive experience interacting with artificial intelligence and machine learning applications on popular social media platforms, putting youth in a unique position to examine, evaluate, and even challenge these applications. Algorithm auditing is a promising candidate for connecting youth's everyday practices in using AI applications with more formal scientific literacies (syncretic designs). In this paper, we analyze high school youth participants' everyday algorithm auditing practices when interacting with generative AI filters on TikTok, revealing thorough and extensive examinations, with youth rapidly testing filters with sophisticated camera variations and facial manipulations to identify filter limitations. In the discussion, we address how these findings can provide a foundation for developing designs that bring together everyday and more formal algorithm auditing.