ON MONDAY, social media firms gave to a US Senate Committee millions of posts, advertisements, and accounts relating to an alleged Russian disinformation campaign targeting US politics. In an agreement with the committee, a group of Oxford University and Graphika Inc researchers had analysed the data. Another report, by cybersecurity firm New Knowledge, too was submitted to the Senate; this is a look at the findings of the Oxford/Graphika report:
Whom it blames
The report says “computational propaganda” was carried out by the Internet Research Agency (IRA), a group of Russians who have been indicted by the US Justice Department. The tactic, first tried in Russian affairs, was reportedly to create false accounts and fake organisations on multiple platforms targeting niche sections of American society with disinformation to polarise them, mostly between 2015 and 2017. The researchers call these actors “cyber troops” who aimed to manipulate public opinion online by using data analytics supplied by social media companies to create micro-targeted silos, or psychological profiling.
The research highlights a multi-platform strategy to spread messages targeting deep social tensions. It says IRA strategy mainly focused on encouraging African-Americans to boycott the election while motivating conservatives to vote for Trump. “Differential messaging to each of these target groups was designed to push and pull them in different ways,” the report states. “The main groups that could challenge Trump were then provided messaging that sought to confuse, distract, and ultimately discourage members from voting.”
More anti-Clinton than pro-Clinton content was posted to liberals, often pitting groups against each other. For example, one post targeting African-Americans stated, “NO LIVES MATTER TO HILLARY CLINTON. ONLY VOTES MATTER TO HILLARY CLINTON.”
The report states the same IP addresses would go so far as to organise opposing protests on two sides of the same street.
Facebook’s Ads Manager allowed IRA to divide these identities into further niches, such as African-Americans interested in the prison system or conservatives interested in policing against African-Americans. Younger viewers were targeted with memes, comedy, and music. White Americans were targeted separately in liberal and conservative segments.
According to the report, IRA spent roughly the same on African-American and conservative messaging but Facebook charged differently for segments. African-Americans, Native-Americans, Latin-Americans, and youth were the cheapest targets; conservatives, Muslim-Americans, and LGBT users were the costliest.
Social media platforms
Instagram saw an increase of IRA activity after the presidential election. Some of IRA’s organisations had PayPal accounts for donations, and their Facebook, Twitter, and Google posts linked to profiles on Medium, Reddit, Tumblr and Pinterest; the report said this allowed IRA to redirect efforts to mediums that had not detected its activities. For example, the report said, Black Matters US had a presence on Twitter, Facebook, Instagram, YouTube, Google+, Tumblr, and PayPal. When their Facebook page was shut down, they re-emerged under “BM” and shifted to themes of black affirmation, used links to redirect traffic, and began purchasing Google Ads with text such as “We are in danger!”
Posts political and general
The report said IRA activity was predominantly “organic posting” — posts by fake pages or accounts masking as a concerned US citizen. This pulls into question the nature of the social media companies’ rebuttal to criticism that they are not doing enough to address these issues. After news of Cambridge Analytica and IRA began to unravel, Facebook, Twitter, and Google accelerated efforts to publicly display advertisements and funding regarding politically or nationally important issues. They began extending their advertisement transparency efforts to countries including India. Instead of only focusing on advertisements, the researchers recommend more research access to all social media content.
From companies’ disclosure to the Senate and public data released by a House Committee, the research found Twitter provided the most data for accounts and posts. Facebook data amounted to about 67,500 organic posts, with almost 31 million shares, almost 39 mn likes, and 3.5 mn comments. Instagram amounted to 116,000 organic posts with 185 mn likes and 4 mn comments. Google supplied the least data, the report said. It did not supply any account data, and disclosed a little over 600 images of ads, 220 YouTube videos, and advertisement data on non-machine-readable PDFs. The report states that Google’s data “was incomplete and provided without context.”