Facebook says hate speech, violent content dropped in Oct-Dec 2020

Facebook on Friday said there’s been a significant drop in hate speech on its platforms, including Instagram, between October and December 2020.

Hate speech views dropped from 10-11 to 7-8 for every 10,000 views of content. The prevalence of violent and graphic content also dropped from 7 views to 5 per 10,000 and adult nudity content dropped from 5-6 to 3-4.

In its latest edition of global Community Standards Enforcement Report, Facebook said it took action on 6.3 million pieces of bullying and harassment, up from 3.5 million in the third quarter of the calendar year. The company attributed the success rate in content removal to advanced technology deployed to detect hateful comments.

Facebook also removed 6.4 million pieces of organized hate content, up from 4 million in Q3. A total of 2.5 million pieces of suicide and self-injury content were taken down, up from 1.3 million in Q3 due to increased ‘reviewer capacity’.

Around 26.9 million pieces of hate speech content were detected and removed due in part to updates technology in Arabic, Spanish and Portuguese, the company said. This is against 22.1 million in Q3.

Facebook’s claims of taking down hate speech, violent, and graphic content come amid an intensifying spat between Twitter and the Indian government over tweets on farmers’ protests. Meanwhile, on its photo-sharing platform Instagram, Facebook cracked the whip on 5 million pieces of bullying and harassment content, up from 2.6 million in Q3. Around 308,000 pieces of organized hate content, up from 224,000 in Q3, were removed in the fourth quarter of 2020. A total of 6.6 million pieces of hate speech content were detected and removed, up from 6.5 million in the previous quarter.

As reviewer capacity increased, Instagram removed 3.4 million pieces of suicide and self-injury content, up from 1.3 million in the third quarter.

Facebook said that the prevalence of such content has come down because of changes made to reduce problematic content in its newsfeed. Each post is ranked by processes that take into account a combination of integrity signals, such as “how likely a piece of content is to violate our policies, as well as signals we receive from people, such as from surveys or actions they take on our platform like hiding or reporting posts” Livemint