Study: leading AI models to detect hate speech are 1.5x more likely to flag tweets by African Americans as offensive or hateful, partly due to social context
and content moderators who grade the test data that teaches these algorithms how to do their job—don't usually know the context of the comments they're reviewing” Then what even is the point of the labeling exercise?! https://twitter.com/... Shirin Ghaffary / @shiringhaffary : A new study shows that AI models are 1.5 times more likely to flag tweets written by black authors as “offensive” compared to other tweets. I wrote about why that is and why it's a problem: https://www.vox.com/... Nima Shirazi / @wideasleepnima : never in a million years could anyone have possibly anticipated this. https://www.vox.com/... Dr. Ziegler / @fakerapper : Black women have been saying this for years. https://twitter.com/...
I am shocked, shocked I say: “two new studies show that AI trained to identify hate speech may actually end up amplifying racial bias” https://www.vox.com/...
This is 100% true..the last two times I've been in twitter jail was for absolute horseshit smh..that algorithm needs to understand context. https://twitter.com/...
Interesting but odd article. It spends about a sentence on the pretty obvious reasons this happens, then explores a narrative practically the opposite of what's going on: The algorithms aren't amplifying people's biases, they're failing to unbias themselves in the ways people do.…
It's no surprise that algorithms are designed with racial bias as a result of lack of representation in Tech: ‘The algorithms that detect hate speech online are biased against black people’ https://www.vox.com/...
“algorithms—and content moderators who grade the test data that teaches these algorithms how to do their job—don't usually know the context of the comments they're reviewing” Then what even is the point of the labeling exercise?! https://twitter.com/...
A new study shows that AI models are 1.5 times more likely to flag tweets written by black authors as “offensive” compared to other tweets. I wrote about why that is and why it's a problem: https://www.vox.com/...