Text-to-Image Bias

Media United States of America

Hugging Face: The Essential AI Toolkit for Journalists and Content Creators JournalistsonHF. The Essential AI Toolkit for Journalists and Content Creators. Hugging Face, n.d., https://huggingface.co/spaces/JournalistsonHF/ai-toolkit. (Accessed March 2025).

The Text-to-Image Bias tool generates multiple images from user prompts to analyze biases in various text-to-image models. It evaluates gender, skin tone, age, and potential sexual nature in generated subjects using models like BLIP for gender detection and Faces Age Detection for age identification. The tool visualizes the data through grids, categorizing attributes such as skin tone, gender, age, and whether the images are appropriate for all audiences (NFAA). This demo provides a critical insight into how biases manifest in AI image generation, aiding in the identification of areas needing improvement.

Newsgathering and Fact-Checking

Data Gathering (Open-source intelligence tools, Geospatial data platforms, Public record datasets) Data Verification (Fact-checking, Reverse-image search, Source tracing) Social Media Mining (Automated filtering, Pattern identification) Hazard Monitoring and Warning Systems

#detect Identify specific attributes such as gender, skin tone, age, and potential sexual nature in generated images. Use models like BLIP and Faces Age Detection for accurate classification.

Website

Developed by: Civil Society
Deployment Type: Web Platform
Community Moderation: Does not require community manager
Difficulty Level: Requires developer
License: Open-source

Back to all

With the support of

User Guide

The designations used and the presentation of materials throughout this repository do not imply the expression of any opinion on the part of UNESCO and IRCAI concerning the legal status of any country, territory, city or area, or of its authorities, or concerning the delimitation of its frontiers or boundaries. The ideas, opinions, and content presented in this repository are those of the authors; they do not necessarily represent the views of UNESCO and IRCAI and do not commit the Organizations.