Facebook took action on 31.5 million pieces of content for hate speech in Q2

Facebook took action against 31.5 million pieces of content for hate speech in the June 2021 quarter, and the prevalence of such content on the platform declined to 5 views for every 10,000 content views globally, the social media giant said in a report.

“We removed 31.5 million pieces of hate speech content from Facebook, compared to 25.2 million in the March quarter, and 9.8 million from Instagram, up from 6.3 million in the first quarter. Prevalence of hate speech on Facebook continued to decrease for the third quarter in a row,” Facebook Vice-President Integrity Guy Rosen said in a blog post.

Hate speech content removal has increased over 15 times on Facebook and Instagram, since Facebook first began reporting it, he added.

Rosen said in the second quarter, the prevalence of hate speech was 0.05%, or 5 views per 10,000 views. This is down from 0.05-0.06%  or 5 to 6 views per 10,000 views in the first quarter of the year.
The statistics are part of Facebook’s Community Standards Enforcement Report for the second quarter of 2021.

Rosen credits AI for hate speech detection

Rosen explained that the reduction was due to “continued improvement in our proactive detection”.
“Our investments in AI (artificial intelligence) enable us to detect more kinds of hate speech violations on Facebook and Instagram. This technology helps us enforce our policies across billions of users and multiple languages,” he said.

Rosen added that steady, continuous AI improvements and advancements, such as the Reinforcement Integrity Optimizer (RIO), enable Facebook’s AI models to spot hate speech using real-world data and improve over time.

During the reported quarter, the company removed more than 20 million pieces of content from Facebook and Instagram globally for violating its policies on Covid-19-related misinformation.

It also removed over 3,000 accounts, pages, and groups for repeatedly violating its rules against spreading Covid-19 and vaccine misinformation.

“We displayed warnings on more than 190 million pieces of Covid-related content on Facebook that our third-party fact-checking partners rated as false, partly false, altered or missing context, collaborating with 80 fact-checking organizations in more than 60 languages around the world,” the company said.

When a piece of content is rated with one of these ratings, a prominent label is added that warns people before they share it, and it is also shown lower in people’s feed.

Besides, the US-based company took action on 2.3 million content pieces on Facebook, and 458,000 pieces on Instagram for child nudity and physical abuse content. Under Child sexual exploitation content, it took action on 25.7 million pieces on Facebook and 1.4 million pieces on Instagram.

During the June 2021 quarter, Facebook took action on 6.2 million pieces of organised hate content (compared to 9.8 million pieces in March quarter), 16.8 million pieces of suicide and self-injury content (as against 5.1 million pieces in the first quarter) and 34.1 million pieces of violent and graphic content (up from 30.1 million pieces in the first quarter).

On Instagram, it took action on 367,000 pieces of organised hate content (up from 325,000 in March quarter), 3 million pieces of suicide and self-injury content (from 2.6 million in the first quarter) and 7.6 million pieces of violent and graphic content (compared to 5.5 million pieces in the first quarter).