Over the years, social media platforms have been trying to train their algorithms to automatically detect and flag content that violates the community guidelines and usage of racist language. Where they have succeeded to a great extent, the algorithms still have a hard time differentiating racist language from the normal conversation at times.
Recently, a YouTube channel that has over a million subscribers and belongs to the famous Croatian chess player Antonio Radic, also known as ‘Agadmator,’ was blocked by YouTube. Although the channel was restored within 24 hours, YouTube did not give a reason why the channel was blocked.
According to experts, the usage of words like “black” and “white” confused Youtube’s AI filters. Researchers have found that 80% of chess videos that were flagged for hate speech were actually reported due to the usage of terms like ‘black,’ ‘white,’ ‘attack’ and ‘threat.’
YouTube uses a combination of AI algorithms and human moderators to filter out prohibited content. Usually, the content goes through the AI algorithm and then is reviewed by the human moderators. In this case, the algorithm couldn’t differentiate between hate speech and normal conversation which is why researchers now suggest that social-media platforms should incorporate chess language in their algorithms to avoid future incidents like this.
This is not the first time Agadmator was blocked. He faced a similar situation in June last year as well when he posted a segment with the Grandmaster Hikaru Nakamura, a five-time champion and the youngest American to earn the title of Grandmaster.
Ashiqur R KhudaBukhsh, a computer scientist at Carnegie Melon’s Language Technologies Institute, opined,
We don’t know what tools YouTube uses, but if they rely on artificial intelligence to detect racist language, this kind of accident can happen.
KhudaBukhsh and his fellow researchers conducted a study where 680,000 comments from five popular YouTube chess channels were analyzed. They found that 82% of the comments flagged in a sample set didn’t include any obvious racist language or hate speech but since they used words like black,’ ‘white,’ ‘attack’ and ‘threat,’ they were flagged.