Inside Facebook's automated abuse algorithm

Facebook should confess / remind / admit that it is next to impossible to become "censor of the internet" and push back some. Stand up for themselves. Having said that however, I am not a fan of their requirement people use their real names: This seems to lead to cases of impersonation: Consider the case of Hanif who's responses from the system said that she was the imposter, because the imposter profile had 4,000 friends and a bulk of pictures - she was unable to have it taken down!

Facebook's abuse algorithm assists it's staff by recognising and filtering duplicate abuse reports in order to more efficiently flag violations such as nudity and pornography, the representative explained. If Facebook receives 1,000 reports on the same thing, the algorithm will streamline the process so humans aren't taking action on all 1,000.

Manually reviewed abuse responses from Facebook should contain a line such as: "Your case has been reviewed by a human. We use form responses due to the high volume of cases we review and use automated systems to assist in our processing of your case. If you wish to speak with a real human, this costs $0.59/minute etc". Provided an option for a paid voice line. This would be an incredibly effective method to enable real peoples request to cut through the spam, since spammers don't like to paid for things.

Stop using your real name

They should cease suggesting people use their real names. I don't, but I was somehow able to change it back in 2015. Probably because I changed mine in the sweet spot between January - when a guy called Phuc Dat Bich had just won the right to use his real name appeared - and November - when it was revealed to be a hoax!

In January 2015, a 23-year-old Australian bank employee claiming to be named Phuc Dat Bich posted a photo of his passport identification page to Facebook, protesting that the company had unfairly shut down his account for being "false and misleading".

Phuc Dat Bich

Phuc Dat Bich

"Is it because I'm Asian? Is it?" he asked. The BBC reported that in Vietnamese the name is pronounced similarly to "Phoo Da Bi". After his reinstatement, Bich posted a thank you note to supportive Facebook fans, stating he was "glad and honoured to be able to make people happy by simply making them laugh at something that appears outrageous and ridiculous". In 25 November 2015 "Phuc Dat" published a further message admitting it was a hoax. The video this BBC page is amusing.

In August 2012, Facebook estimated that more than 83 million Facebook accounts are fake accounts. As a result of this revelation, the share price of Facebook dropped below $20.

In 2008, a woman in Japan named Hiroko Yoda had her Facebook account suspended over her surname somehow being confused with Master Yoda from Star Wars fame. In USA, a native with last name "Kills The Enemy" ran into trouble: Robin Kills The Enemy is a resident of the Sioux Rosebud reservation in South Dakota had been using "KillsTheEnemy"as a last name up until she contacted Facebook to try and change it to the regular spelling with spaces.

In 2015, a Gaelic-speaking retired policeman won the right to use his Gaelic name on Facebook, even though it was not his legal name. Some Scottish Gaelic surnames, such as NicIllAnndrais, contain more than two capital letters and are still disallowed by Facebook (according to Wikipedia):

The Facebook real-name policy controversy refers to the controversy over social networking site Facebook's "real-name system" dictating how people register their accounts and configure their user profiles. The controversy stems from a policy that those who have been adversely affected describe as penalising users who are in fact using their real names which Facebook has nevertheless deemed to be "fake", while simultaneously allowing anyone to create fake yet plausible-sounding names, as well as obviously implausible-sounding names comprising word combinations that Facebook's software fails to recognise as unlikely to be real because their alleged nonstandard spelling or confusion with fictional characters. Facebook furthermore prohibits users from accurately representing names which according to the site have "too many words", and prohibits initialising first names, preventing users who do so in real life from formatting their own names as they see fit.

Food for thought. Maybe I should add a Facebook comments feature to my blog. Or then again, maybe not.

Fake News

Facebook terms show the use of algorithm to adjust "post distribution", I believe this is called EdgeRank, and it's formula to show if someones post will appear on your screen is roughly: FriendAffinity * MediaItemURLPopularity / TimeSincePosted. 

Reducing the spread of false news on Facebook is a responsibility that we take seriously. We also recognise that this is a challenging and sensitive issue. We want to help people stay informed without stifling productive public discourse. There is also a fine line between false news and satire or opinion. For these reasons, we don't remove false news from Facebook, but instead significantly reduce its distribution by showing it lower in the News Feed.

Violence

They do a good job here. The titles of the section are shown below the only I quote was the first one:

  • Credible violence - We aim to prevent potential real-world harm that may be related to content on Facebook. We understand that people commonly express disdain or disagreement by threatening or calling for violence in facetious and non-serious ways. That's why we try to consider the language, context and details in order to distinguish casual statements from content that constitutes a credible threat to public or personal safety. In determining whether a threat is credible, we may also consider additional information such as a targeted person's public visibility and vulnerability. We remove content, disable accounts and work with law enforcement when we believe that there is a genuine risk of physical harm or direct threats to public safety.
  • Dangerous individuals and organisations
  • Terrorist activity
  • Organised hate
  • Mass or serial murder
  • Human trafficking
  • Organised violence or criminal activity
  • Promoting or publicising crime
  • Coordinating harm (org crime)
  • Regulated goods
Posted by tomachi on March 28th, 2019 filed in Technology