,
7 tweets,
3 min read
Read on Twitter
I'm sharing some scholarly research on facial recognition w/ @detroitpolice to continue the dialogue. While facial recognition identifies a white man’s gender 99% of the time, it misidentifies darker-skinned females up to 35% of the time. nytimes.com/2018/02/09/tec… 1/7
Another study in 2018 from the National Institute of Standards and Technology found that facial recognition falsely matches Black women’s faces 10 times more frequently than white women. This is a deeply flawed technology. wired.com/story/best-alg… 2/7
Because this technology is untested, it will rely on real people to make judgments about whether someone is a match. Scientific research about the cross-race effect shows that people are less accurate in judging faces of people of a different race. ncbi.nlm.nih.gov/pmc/articles/P… 3/7
A Northwestern University study showed that we have better memories of faces of the same race because we’re more perceptive of same-race faces, and that we are less likely to recognize or distinguish between other-race faces. frontiersin.org/articles/10.33… 4/7
A study at the University of Glasgow suggests that our brains encode other-race faces primarily according to the racial group they belong to, rather than by distinguishing features. pnas.org/content/107/46… 5/7
Even Chief Craig admitted this technology isn’t perfect & the human element plays a major role. The science suggests that people have difficulty identifying faces of people from other races & we must take it seriously. I hope @detroitpolice will read these studies. 6/7
At the end of the day, I was elected to serve my residents, and I cannot in good conscience sit by while inaccurate facial recognition technology is deployed in ways that run the risk of false arrests and overpolicing. We need a real dialogue here. 7/7
