As a Black woman, I’ve experienced what it has felt like to be seen as a monolith. If other Black women are into something, then I must have to be into that as well… right? Whether it is the music we listen to, the outfits we wear, or the way we do our makeup, it always seems as if Black women are put into these stereotypical boxes of how we should present ourselves to the world. This is why I wanted to see how AI views Black women. Will it do the same things? Put Black women into a stereotyped box? Or will it show a more diverse depiction of Black women?
Using, https://imggen.ai/, I typed in, “A group of Black women”, and had it generate an image. To my surprise, the image staring back at me didn’t feature Black women with Eurocentric features (straight hair, a thin nose, light complexion). Instead, I looked at a group of Black women with natural hair and protective styles, women with different complexions, and a wider nose bridge. However, unsurprisingly, it only shows one type of Black woman, the African woman.
These women are clad in dresses akin to traditional African fabrics and designs, such as kente cloth. They all have features that resemble women from African countries, as well. So, I asked AI to repeat this, but this time I simply typed in “Black women” and again, all the images generated reflect Black women from the African diaspora. Is this how AI sees Black women? As only being African women? Not only that but does AI think being an African woman is this homogenous?
It is not just me asking these questions, in a New York Times article, many Black artists are doing the same questioning, particularly Stephanie Dinkins, a Brooklyn-based artist. For several years, Dinkins has been experimenting with AI’s ability to accurately depict Black women. She started by asking AI to generate images of Black women smiling and found a lackluster response.
Zachary Small, the journalist who wrote this article, reported, “Her algorithm produced a pink-shaded humanoid shrouded by a black cloak”. Dinkins was expecting to see something that resembled Black womanhood, but instead, all her results when generating images of “African American woman” or “Black woman,” came back distorted with mangled facial features and hair textures.
In the same article, Senegalese artist Linda Dounia Rebeiz, experienced something similar to me. When using AI to provide images of her hometown, Dakar, Rebeiz found, “The algorithm produced arid desert landscapes and ruined buildings…” which she believes is how the West views Africa.
While I did not get distorted or negative depictions of Black women, I very much saw how AI reflects how Western countries view Black people. It comes off as if we are a monolith from one region in Africa. I feel as though the view of Black people being a monolith is constantly reinforced and normalized in the United States, and I feel as though it has a lot to do with the language used to describe Black people.
Cyndey Adams breaks it down perfectly in her article, “many people often default to “African American” out of a desire for either political correctness or politeness. [“Black” and “African-American”] are often used interchangeably, but that isn’t always accurate, and it’s important to understand the nuance when discussing race both in America and on a global scale”.
Blackness is not limited to just Africa, as Blackness can be found around the globe, and I believe that having that basic understanding can help expand how non-Black people perceive us in AI creation. However, it makes me relieved to know that Black women are becoming more involved with the evolution of AI. While it’s easy to believe that any representation is great, it’s important to remember that accurate and inclusive representation makes a difference, especially to communities of color.