https://www.google.com/amp/s/www.vo...-bias-black-african-american-facebook-twitter
If you've paid attention to multiple social media outlets. You would know that they use algorithms to monitor peoples post. Sites like Twiiter, FB, and YT can have so many users and post this is almost impossible to be done by human moderators. So they use algorithms to search for hate speech and other offensive language they don't want promoted on their site.
Problem is, the algorithms are color blind. It can tell the difference between a white person using the n-word and a black person using it. So as stated in the article above black folk have been getting flagged and banned for hate speech at up to and in some cases more than 1.5x than white folk.
Even when the algorithms have been modified to take the posters ethnicity and background into effect it only reduced the number of flagged post by black folk by 11%.
Recently some LGBTQ+ content creators on YT filed a lawsuit on Google for discrimination. Because their algorithms flagged sexual content and words. The YouTibe algorithm scans the first 10 mins of a video for words spoken. Any sexual references, or hate related speech instantly demonized and down ranks that video. Even just showing a picture of Hitler and calling him a monster will get you demonitized and possibly banned for hate speech.
https://www.google.com/amp/s/www.th...-alleged-video-recommendations-demonetization
So what do you think? Do we ride out the rough spots until AI is advanced enough to take speech into context or should we stop using it and just go back to human moderators?
If you've paid attention to multiple social media outlets. You would know that they use algorithms to monitor peoples post. Sites like Twiiter, FB, and YT can have so many users and post this is almost impossible to be done by human moderators. So they use algorithms to search for hate speech and other offensive language they don't want promoted on their site.
Problem is, the algorithms are color blind. It can tell the difference between a white person using the n-word and a black person using it. So as stated in the article above black folk have been getting flagged and banned for hate speech at up to and in some cases more than 1.5x than white folk.
Even when the algorithms have been modified to take the posters ethnicity and background into effect it only reduced the number of flagged post by black folk by 11%.
Recently some LGBTQ+ content creators on YT filed a lawsuit on Google for discrimination. Because their algorithms flagged sexual content and words. The YouTibe algorithm scans the first 10 mins of a video for words spoken. Any sexual references, or hate related speech instantly demonized and down ranks that video. Even just showing a picture of Hitler and calling him a monster will get you demonitized and possibly banned for hate speech.
https://www.google.com/amp/s/www.th...-alleged-video-recommendations-demonetization
So what do you think? Do we ride out the rough spots until AI is advanced enough to take speech into context or should we stop using it and just go back to human moderators?
