
Adobe Obtains Patent for AI Diversity Auditing Software
By Movieguide® Contributor
Adobe, a computer software company, has just been issued a patent for image-scanning software that audits photos to identify and categorize diversity.
After sifting through images, “the system then calculates a ‘diversity score’ for the set of images using machine learning by comparing the classified images to the comparison population,” the Daily Upside reported. “Finally, the system then ‘augments the set of images to increase diversity’ using ‘additional retrieved images’ until a certain threshold of diversity is met.”
This software can only evaluate images. If a user searched for what Adobe calls “sensitive attributes such as race, age, or gender” (Patent No. 20230267764) and used it to classify people’s images based on race, it would only group those who appear to be of a certain race. For example, it could group together images of people who appear white, but it would not be able to tell if they are truly, factually Caucasian.
Companies could use this software to categorize their employees or to see how many people from certain groups attended an event. But the intent for the program’s creation is a little different.
The Daily Wire wrote, “The inventors—Mehrab Tanjim, Ritwik Sinha, Moumita Sinha, David Thomas Arbour and Sridhar Mahadevan—published the paper, ‘Generating and Controlling Diversity in Image Search,’ declaring that ‘generations of systemic biases’ resulted in racial and gender demographic patterns existing in certain professions.”
“The pursuit of a utopian world demands providing content users with an opportunity to present any profession with diverse racial and gender characteristics,” stated the report. “To remedy these problems, we propose a new task of high-fidelity image generation by controlling multiple attributes from imbalanced datasets.”
“The researchers also stated that stock image and image search engines reflect that bias, and that AI-generated images would be necessary because algorithm tweaking wouldn’t go far enough to correct the issue,” the Daily Wire reported.
This is where image augmentation comes into play. The system would modify the results so that a user or image search tool could achieve the desired or predetermined scale of diversity.