I wanted to continue to explore bias in our algorithms through machine learning. This art project is merely a comment on what I discovered, not a scientific experiment.
For this project, I downloaded a library of images off of Google Images using the words ‘boss’ and ‘bitch.’ I then ran my image data through a machine learning algorithm using TensorFlow to train my machine to recognize these images. After that, I downloaded an image library for ‘men’ and ‘women,’ merged the library, and asked the machine to filter these images and assess it based on the questions: “What images are the most ‘boss’”? and “What images are the most ‘bitch’”? The machine identified the images and gave it a % prediction based on confidence levels. I took the average of the top 30 and created a “generative” image of my results. I created two portraits of ‘boss’ and ‘bitch,’ one spanning from (July ’01-July ’10) and the other, (July ’01-present).
July 2001 – July 2010
July 2001 – Sept 2018
