ARTIFICIAL INTELLIGENCE
Joy
Buolamwini
FOUNDER /
Algorithmic Justice League
Calling out bias
hidden in tech tools.
IN 2015, MIT Media Lab master's student Joy Buolam-
wini gazed, puzzled, into her webcam. She was work-
ing on an art piece to project digital masks on faces, but
her own was invisible to the off-the-shelf face-tracking
software she was using. She scrawled a face on her
hand in marker. Face detected. Hmm. She covered her
dark skin with a white mask left over from Halloween.
Face detected. Oh.
Buolamwini had come to MIT to explore how people
could use technology for social change. The algorithm
that couldn’t see her face set her on a new path: explor-
ing how tech could be misused and abused. When she
learned how facial recognition is used in law enforce-
ment, where error-prone algorithms could have grave
consequences, she says, “that’s when it became urgent.”
Buolamwini, now working on a PhD at MIT, is among
those pressing companies and governments to be more
cautious in their embrace of AI. Her work has revealedbiases in facial-analysis algorithms from Amazon, IBM,
and Microsoft. She showed that the services frequently
saw women with dark skin as men but made few errors
on men with light skin. (All three companies say their
tech is now more accurate.)
Buolamwini believes that people developing AI tech
should check their inventions for bias and other harms
before launch, not rely on Good Samaritans like her to
audit them later. She spearheaded the Safe Face Pledge
with Georgetown Law—a call for companies to agree
to take steps aimed at mitigating the harms of facial
analysis. She’s also pushing for a moratorium on police
and government use of facial recognition to allow time
for debate and possibly the creation of regulations. “We
need public deliberation,” Buolamwini says. Her own
deliberation takes place in scientific papers, congressio-
nal testimony, and spoken word poetry. “It’s our stories
that lead to change,” she says. —TOM SIMONITE