Propaganda on Facebook: AI defeats the “war machines”

Social media erupted late on April Fool’s Day after Facebook removed pages from its site over what it called “coordinated inauthentic behavior and spam from India and Pakistan”.

The pages removed (unpublished, in technical language) included ones with millions of followers, which supported the Bharatiya Janata Party government: TheIndiaEye, NationWantsNaMo, MyNation, DainikBharat, and Postcard. These pages are not owned by the BJP and they do not take instructions from the party. But almost all the content in these pages promote the BJP and Prime Minister Narenda Modi, and vehemently, and oftentimes vulgarly, criticise the Congress and its leaders.

In itself, this does not mean anything, and the pages could have been in operation. But it is the coordination in behaviour, which means the same content (videos, pictures, audio and written word) is uploaded into multiple pages, that makes the Facebook algorithm notice these posts. 

The BJP has been at the forefront of occupying cyberspace across social media platforms, a strategy that paid it rich dividends in 2014 and later in Assembly elections across the country.

The Congress, a late starter, had 687 pages which had been taken down. These pages did not even have a tenth of the following that the BJP-liked pages had.

In a release, Nathaniel Gleicher, Head of Cybersecurity Policy at Facebook, said: “[Facebook] removed Pages, Groups and accounts for violating Facebook’s policies on coordinated inauthentic behavior or spam. Today’s action includes four separate takedowns—each distinct and unconnected.

“We removed 103 Pages, Groups and accounts on both Facebook and Instagram for engaging in coordinated inauthentic behaviour as part of a network that originated in Pakistan.

“We removed 687 Facebook Pages and accounts—the majority of which had already been detected and suspended by our automated systems—that engaged in coordinated inauthentic behavior in India and were linked to individuals associated with an IT Cell of the Indian National Congress (INC).

“We removed 15 Facebook Pages, Groups and accounts that engaged in coordinated inauthentic behavior in India and were linked to individuals associated with an Indian IT firm, Silver Touch.

“We removed 321 Facebook Pages and accounts in India that have broken our rules against spam. Unlike the first three actions, this last activity does not represent a single or coordinated operation — instead, these are multiple sets of Pages and accounts that behaved similarly and violated our policies.”

Facebook used machine learning, data analytics, and artificial intelligence to take down the pages, and this was not about fake news but something much bigger in scope and which fell in a grey area of authenticity.

Venkat Ananth, a journalist who has written on the subject, explains in his Twitter post: “What Facebook’s automated system or AI algorithm has ‘likely’ detected is this cross posting behavior, specifically to drive engagement. It [AI] doesn’t likely know what they’re posting, because that’s not even in contention. It is looking for behavior, by ‘likely’ matching the same content across multiple pages, of whom admins have already identified. That behavior, in Facebook parlance could be “inauthentic”.

The behaviour violation, Venkat explains, could be “posting massive amounts of content across a network of Groups and pages in order to drive traffic to websites they are affiliated with in order to make money”. In short, this is not fake news but a behaviour pattern that violates FB policy.    

Asked to differentiate between fake news and this behaviour, Amantha Perera, Asia Pacific Coordinator, DART Center for Journalism and Trauma, Graduate School of Journalism, Columbia University (based in Sri Lanka), told Frontline, “Fake news is lies, or something that is bordering on lies, which is spread with a certain motive in mind. It could be to change attitudes and opinions, which is most often the case, but there could be other motives as well, even monetary. 

“Behaviour here is the manner in which people use content, which is not necessarily fake news. It could be a one  hundred per cent authentic piece of information, but someone with control of 20 FB groups with access to 2million users can post it on all those pages, and repost and re-repost. It could also be, in a more simplistic way, someone who keeps bad mouthing someone using racial and sexist slurs online. Behaviour is how we deal with the content.”