Facebook has doubled its bullying-related removals
Facebook released a new Transparency Report that shows a significant rise in removals of bullying, which peaked at 6.3 million removals during the fourth quarter of 2020.
These numbers are an increase from 3.5 million in the last quarter and 2.8 million in the fourth quarter of 2019.
Much of the change is due to improvements in the automated systems that analyze Facebook and Instagram comments, the company said.
The latest Facebook transparency report covers the period from October to December 2020, which includes the US presidential election.
The main Facebook network during that time removed more harassment, organized hate speech, suicide and self-harm content.
Instagram has also seen huge leaps in the removal of bullying.
The company says its numbers are shaped by two factors: more human review capacity and improvements in artificial intelligence, especially for non-English language participants.
The company also indicates that it is relying on automation to process the growing amount of video and audio across its platforms.
We are investing in technology in all kinds of different ways, we understand audio and video, we understand the content related to these things, and who has shared them, and we build a broader picture of what is happening there, Facebook said.
“Obviously, we are looking at how products and monetization are changing before those changes to make sure we have the technological tools we need,” she added.
The increased oversight helped Facebook and Instagram remove more suicide and self-harm posts.
Facebook removed 2.5 million offending content, compared to 1.3 million in the previous quarter, while Instagram removed 3.4 million, up from 1.3 million.
Facebook attributes some of the increases to AI-powered moderation, as it removed 6.3 million bullying and harassment content across Facebook, nearly double the numbers from previous quarters.
Meanwhile, AI-powered moderation via Instagram removed 5 million of content, up from 2.6 million in the last quarter and 1.5 million at the end of 2019.
These increases stem from technology that better analyzes comments in the context of the accompanying post.
Supervision across non-English languages was a historical weakness for Facebook, but it improved AI discovery in Arabic, Spanish and Portuguese, increasing hate speech removal from 22.1 million to 26.9 million.
However, this is not as big as the leap that Facebook experienced in late 2019, when it introduced what it called drastic improvements to automatic detection.
Facebook said it changed its news feed in ways that reduce the amount of hate speech and violent content people see.
The company said it was still formulating responses to some of the suggestions from Facebook's supervisory board, which issued its first decisions last month.
As it did in the last quarter, Facebook suggested that lawmakers could use its transparency report as a model for the legal framework.
Facebook has supported changes to Section 230 of the Communications Etiquette Act, the broad liability shield that has come under attack from social media critics.