Blog:Bias in Artificial Intelligence Confirmed
What we've suspected
Last month, a research piece in MIT News was published confirming what some folks have suspected through current experiences with face-recognition "solutions" thus far: they don't recognize us—people of color. Disappointing; not surprised. Effectively, the research found that there is gender and skin-type bias in commercial artificial-intelligence systems.
Curious, Buolamwini, who is black, began submitting photos of herself to commercial facial-recognition programs. In several cases, the programs failed to recognize the photos as featuring a human face at all.
One might ask the question, "What's the big deal?" Well, this can have (and already has) adverse impact on individuals' lives, especially the already marginalized. Adverse effects can range from day-to-day interactions, like someone wanting to use a photo tool to help tag and group their friends and family but can't rely on the tool properly because they and their circle of friends and family have darker skin, to, someone being wrongfully identified using biased face-recognition software in a legal matter.
Representation. It matters.
This is why
representation matters: it matters in big tech and across all of society, e.g. elected positions, professional leadership positions, in the news, in the classroom, etc.
The data set used to assess its performance was more than 77 percent male and more than 83 percent white.
Increase access to education and personal growth
To help address the lack of representation across industries, we—society—can begin by removing barriers-of-entry to underrepresented communities, starting by making higher education accessible to everyone and valuing alternative forms of higher learning, like vocational/trades. Limiting access to education only serves to keep the current power structure as-is, which is only favorable to a small subset of people when we consider everyone in our communities.
“To fail [...] on something that’s been reduced to a binary classification task, you have to ask, would that have been permitted if those failure rates were in a different subgroup?” Buolamwini says.
Reform laws based on codified discrimination
Equally important, existing policies and laws need to be reviewed regularly and rewritten with an inclusive, intersectional lens to address any and all forms of discrimination. Spoiler: there's a lot of discrimination already codified in the books at all-levels of government!
- Timnit Gebru (Researcher)
- Joy Buolamwini (Researcher)
- Research Publication Source (MIT News)
- Inclusive Design
- Product Philosophy
I operate from a place of compassion, possibility and imagination. My work and efforts share a common goal: create a better, sustainable and equitable world by building inclusive communities, products & experiences.