Facebook removed a staggering 2.5 million examples of “hate speech” from its platform in the first quarter of 2018, according to data released by the social network.
Facebook, which like Google is currently working on A.I. technology to identify “hate speech”, admitted that its current hate-speech-detection technology “still doesn’t work that well” and that automatically flagged content “needs to be checked by our review teams”.
Nevertheless, Facebook said that it “removed 2.5 million pieces of hate speech in Q1 2018”, and that 38 percent of it was flagged by the social network’s automated systems.
The hate speech numbers were released as part of Facebook’s first “Community Standards Enforcement Report”, which contains aggregate data on some of Facebook’s enforcement actions.
Absent from the report is a precise number of how many accounts and pages have been banned or suspended from the platform over “hate speech”, and how many have been mistakenly suspended and then restored. Facebook often mistakenly suspends the accounts of prominent conservatives only to restore them later — the most recent example being Islam critic Pamela Geller.
Also contained in the report:
- Facebook banned “about 583 million fake accounts — most of which were disabled within minutes of registration.” (note: this is not the same as monthly active users.)
- Facebook took down approximately 3.5 million pieces of violent content in Q1 2018, 86 percent of which was automatically detected by the social network’s technology.
- Facebook took approximately 21 million pieces of adult nudity and sexual activity in Q1 2018, 96 percent of which was automatically detected.
- Facebook took down 837 million pieces of spam in Q1 2018.
Facebook says the release of these numbers is part of its commitment to transparency. But, as noted above, it has not released data on how many accounts were mistakenly suspended, particularly for hate speech. Nor has the tech giant followed Trump campaign manager Brad Parscale’s suggestion that Facebook opens up its lists of banned content for independent oversight.
Facebook has admitted that its systems for automatically detecting “hate speech” don’t “work that well,” and Mark Zuckerberg himself was unable to provide a definition for “hate speech” when questioned on the subject by a Senate committee. Nevertheless, the Facebook CEO says he aims to have a functioning A.I. system to root out “hate speech” on the platform within 10 years.