The “violative view rate” (VVR) has dropped over 70% since it was first tracked in the fourth quarter of 2017, YouTube said, and demonstrates its progress in blocking hate speech and other videos it considers dangerous before they go viral.
Liking this story?
Get one mail covering top tech news of the day in under 5 minutes!
Critics have said inadequate policing by YouTube and other social media companies enables false and hateful rhetoric to spread, fomenting deadly violence such as the US Capitol attack in January.
YouTube’s VVR was steady over the last six quarters measured, according to the new data, which run through 2020. The VVR comprises all policy violations and is derived from a sampling of videos. It does not include comments on videos.
Jennifer O’Connor, a product director at YouTube, told reporters that she hoped releasing the estimate each quarter “continues to hold us accountable”. She said the rate, like other enforcement data YouTube releases, could fluctuate as its technology, rules and users evolve. For instance, YouTube removed nearly 171,000 channels for hate speech in the fourth quarter, three times more than the preceding period. It attributed the jump to improved detection technology.
Facebook Inc. releases a similar estimate but excludes bullying, spam and other violations. Added up, Facebook has said at least 15 million views out of 1 billion in the fourth quarter were of content violating its rules against adult nudity and sexual activity, violent or graphic material and hate speech. Countering criticism about “grading” itself, Facebook last year said it would hire an outside auditor to assess its disclosures.
YouTube’s O’Connor on Monday declined to commit to an external audit but said she “wouldn’t rule it out”.