A Case for AI Literacy: Generated Images Reproducing Biases

• In Event: Virtual Poster Session
  In Virtual iPresentation Only: SIG-Technology as an Agent of Change in Teaching and Learning Virtual Poster Session

Abstract

In this paper, we describe a study of four popular word games that use AI-generated images for their clues. We used content analysis and a statistical parity difference (SPD) model to analyze 182 images. We found that overall, all the games favored representations of white male figures, although some were less biased than others. We offer this study as one practical example of how teachers might offer their students an opportunity to develop AI literacy. Underlying this specific study are key concepts teachers could focus on: guiding students to evaluate AI tools, using their lived experiences, systematically collecting data, identifying and applying appropriate analysis techniques, and producing their own insights and understanding.

Authors

• Matthew Duvall, University of Pennsylvania

• Chen Wang, University of Pennsylvania