Search algorithms continue to oppress

Updated: 2 days ago



In February 2019, I gathered in a jam packed hotel ballroom in downtown Edmonton to hear Dr. Safiya Noble talk about her book Algorithms of Oppression: How Search Engines Reinforce Racism. This was early days for me in my study of AI ethics and learning about how discrimination can be baked into the technological tools we use. I was as shocked and disturbed as the hundreds of other people in the room as we listened to Dr. Noble's stories about conducting research on Google, uncovering racist, sexist imagery and text in the course of her work. This was my first introduction to the idea that search is not neutral nor objective; that value systems are at play which perpetuate a culture of discrimination on these platforms. Dr. Noble's book became an important resource for me in my research.

Recently, I found myself having a "Safiya Noble moment" as I stumbled across some racist search results while doing work for a client.

I'm consulting on a podcast about anti-Black racism. I was trying to pull together a quick prototype website for my client, so I went to Weebly to put together a free site. I wanted to show my client some images of Black people. I decided to use the built in photo search to look for "black people" and I got this....

Image: Weebly, photos from Flickr - Retrieved on February 11th

I was initially taken aback - am I doing this right? Thinking maybe I needed to be more specific, and also to maybe capitalize Black, I tried "Black woman". I got the same result. I then tried "Black man".

Image: Weebly, photos from Flickr - Retrieved on February 11th

I've been working in the AI ethics space for two years now. I know what's going on in terms of how this happens....and yet....having it happen in 2021, after everything that took place last year - George Floyd, Black Lives Matter, EDI promises, the work of people like Dr. Noble and others in our AI ethic community...I was still kind of shocked and horrified.

There is a difference between knowing something in theory, as a researcher, but then seeing it actually happen in real life. I suppose I thought that in this moment of heightened awareness around anti-racism, that companies would actively try to do better. The fact that this imagery was so...blatant...it stunned me.

I tweeted this at Weebly and they responded by saying this was a Flickr issue. I did two more searches - call this my control group - one for White woman and one for White man...

Image: Weebly, photos from Flickr - Retrieved on February 11th

Image: Weebly, photos from Flickr - Retrieved on February 11th

In summary, I got racist images that were not even of people while searching for Black people and Black women (they were invisible), gangster stereotypes for my search of Black men, sexist imagery for White women and a professional looking person in a suit while searching for a White man. Sadly, that sounds about right.

-- Katrina Ingram


#AIEthics #CodedBias #RacistAlgorithms #BLM

_____________

Resources

Dr. Noble recorded an interview with a local podcast while in Edmonton. Learn more about her work.

-- Katrina Ingram