 |
| Facebook has doubled down on bullying-related deletions |
Facebook released a new transparency report that found that bullying deletions have increased dramatically, peaking at 6.3 million removals in the fourth quarter of 2020.
These numbers represent an increase from 3.5 million in the previous quarter and 2.8 million in the fourth quarter of 2019.
The company said most of the changes were due to improvements in the automated systems that analyze Facebook and Instagram comments.
Facebook's most recent transparency report covers the period from October 2020 to December 2020, including the US presidential election.
Meanwhile, the main Facebook network eliminated further harassment, organized hate speech, suicide and self-harm.
Instagram has also taken a big step forward in eliminating bullying.
The company said its numbers were affected by two factors: increased human oversight and improved artificial intelligence, especially for non-English speakers.
The company also said it is relying on automation to handle the increasing amount of video and audio on its platform.
Facebook said we are investing in technology in different ways. We understand audio and video, we understand what these things are about and who is sharing the content, and we can have a broader understanding of what's going on there.
She added, "Of course we are looking at how to change products and generate income before these changes to make sure we have the technical tools we need."
More oversight will help Facebook and Instagram suppress more information about suicide and self-harm.
Facebook removed 2.5 million violating content, up from 1.3 million in the previous quarter, while Instagram removed 3.4 million, down from 1.3 million in the previous quarter.
Facebook attributed some of the growth to AI-based moderation, with 6.3 million Facebook bullying and harassment content eliminated, double what it was in previous quarters.
Meanwhile, thanks to AI-powered verification on Instagram, 5 million pieces of content have been removed, up from 2.6 million in the previous quarter and 1.5 million in late 2019.
These increases are a result of the techniques used to better analyze comments in the accompanying articles.
Monitoring non-English languages is a historical weakness of Facebook, but it improved AI in Arabic, Spanish and Portuguese and increased hate speech elimination from 22.1 million to 26.9 million.
However, it's not a big deal like what Facebook went through in late 2019 when Facebook introduced a major improvement in auto-detection.
Facebook said it changed its news sources to reduce hate speech and violent content people see.
The company is still responding to some recommendations from Facebook's board of directors, which published its first decision last month.
As in the previous quarter, Facebook recommended that the legislature use its Transparency Report as a template for the legal framework.
Facebook supports the Section 230 Amendment to the Communication Etiquette Act, a broad liability shield attacked by social media critics.