How Facebook’s AI Can Catch 95% of the Hate Speech Circulating the Platform

How Facebook’s AI Can Catch 95% of the Hate Speech Circulating the Platform

Spread the love

Recently, Facebook revealed that its AI algorithm is taking strong steps in mitigating the occurrence of hate speech on its platform. The AI algorithm working at the backend of the platform has been successful in removing 95% of the hate speech comments, posts and shares before anyone reports. Facebook revealed this information in its latest community-standard enforcement report explaining 

how the remaining 5% of the 22 million flagged posts were the only shares reported by users. 

According to the published report, there’s also a new hate-speech metrics called the prevalence. Prevalence is an equivalent term used to measure the percent of time someone observes something as harmful. It is done by sampling the content on Facebook based on the count of views and then applying a label on it that it shouldn’t be there. As per the record, the figure was between 0.1 percent & 0.11 percent which approximated around 10-11 views of every 10,000. 

Mike Schroepfer, who is hired as the Facebook’s chief technology officer (CTO), explained this figure in a blog post where he explained that the numbers increased from 80.5% from last year to 94.6%. In 2017, the figure of hate speech recognition was around 24% which was shared in Facebook’s latest Community Standards Enforcement Report. Many different social media firms such as Twitter, Facebook & TikTok have done seemingly well in keeping racial slurs & religious attacks off their platform. Today, companies employ over thousands of content moderators from around the world to help observe different posts, videos and photos which are abundantly shared. 

More than 200 content moderators emailed the CEO Mark Zuckerberg recently explaining their complaints on how they had to risk their lives by joining the office and continue work during the COVID

With the passage of time, it was pretty evident that human content moderators for Facebook weren’t good enough. With the advent of AI and Machine Learning, Facebook made due investments. With powerful machine learning algorithms at work, the content moderation experienced significantly improved. As Mr. Schroepfer states, 

“A central focus of Facebook’s AI efforts is deploying cutting-edge machine learning technology to protect people from harmful content, with billions of people using our platforms, we rely on AI to scale our content review work and automate decisions when possible. Our goal is to spot hate speech, misinformation, and other forms of policy-violating content quickly and accurately, for every form of content, and for every language and community around the world.”

For now the algorithm is in its nascent stage. There are many different things such as sarcasm and image text which Facebook’s algorithm is unable to clearly read or detect. If there are people who are using slang language, this is something which the algorithm was unable to detect but the human moderators could. So until the AI matures, human moderators will never become obsolete. 

For now, the two technologies working at the back-end for detection are RIO and Linformer as concluded by Schroepfer. These tools are responsible for analyzing Facebook & Instagram content.  

Furthermore, Facebook has developed a new tool to detect deep fakes (computer-generated videos that appear very real.) It has made some significant improvements in an existing tool called the SimSearchNet which is an image-matching tool which can help spot misinformation. 

Schroepfer further concludes, 

“Taken together, all these innovations mean our AI systems have a deeper, broader understanding of content,” said Schroepfer. “They are more attuned to things people share on our platforms right now, so they can adapt quicker when a new meme or photo emerges and spreads.”

For now Facebook insists on taking down at least 95% of the overall hate speech proactively. During the last three months, the company has been encouraging employees to take proactive actions. 

Are you encountering incidents of hate speech on the Internet? Best that you feel free to share some of the insights with us. At Tech Wire Hub, we would love to hear back important things from you. 

Leave a Reply

Your email address will not be published. Required fields are marked *