Meta on Wednesday released its biannual Transparency Reports for the first half of 2022. The reports noted that the prevalence of bullying and harassment-related content remained low at 0.08% on Facebook and between 0.04-0.05% on Instagram.
(For insights on emerging themes at the intersection of technology, business and policy, subscribe to our tech newsletter Today’s Cache.)
Meta’s actions against hate speech-related content decreased from 13.5 million to 10.6 million in Q3 2022 on Facebook because of the improved accuracy of its AI technology, according to the company.
“We’ve done this by leveraging data from past user appeals to identify posts that could have been removed by mistake without appropriate cultural context,” wrote Guy Rosen, VP of Integrity.
Now, its AI can better recognise funny terms of endearment used between friends, or better detect words that may be considered offensive or inappropriate in one context but not another, the reports say. “As we improved this accuracy, our proactive detection rate for hate speech also decreased from 95.6% to 90.2% in Q3 2022.”
Similarly, actions against content that incites violence decreased from 19.3 million to 14.4 million in Q3 2022 after improving the AI technology. It was able to better recognise language and emojis used in jest between friends.
“As we improved our accuracy on this front, our proactive rate for actioning this content decreased from 98.2% to 94.3% in Q3 2022,” the reports mention.
For bullying and harassment-related content, the proactive rate decreased in Q3 2022 from 76.7% to 67.8% on Facebook, and from 87.4% to 84.3% on Instagram.
“This decrease was due to improved accuracy in our technologies (and a bug in our system that is now resolved),” the report said.
Although the proactive rate dropped in a few areas, prevalence of harmful content on Facebook and Instagram remained relatively consistent between Q2 and Q3.
In Q3 2022, Facebook took action against 16.7 million pieces of content related to terrorism, an increase from 13.5 million in Q2, as well as 4.1 million pieces of drug-linked content, an increase from 3.9 million in Q2 2022, and 1.4 billion pieces of spam content, an increase from 734 million in Q2.
Similarly, Instagram took action against 2.2 million pieces of content related to terrorism, up from 1.9 million in Q2, and against 2.5 million pieces of drug-linked content, an increase from 1.9 million.
Recently, Meta announced a community forum taking place in December, which will bring together nearly 6,000 people from 32 countries to discuss conduct in the metaverse.