4

This article, which is being linked and shared on Facebook, claims that British MP and Brexit campaigner Jacob Rees-Mogg is receiving support on social media from known Kremlin networks.

Analysis of social media traffic between January and February 2018 shows support for Jacob Rees-Mogg is coming from known Kremlin networks. ...

With a network of accounts pushing Kremlin propaganda from George Soros memes to fake news around EU immigration, Rees-Mogg is now firmly embedded as a popular figure for Russia to champion and its disinformation network is using the botnet to create large numbers of impressions using retweets to push the narrative into the mainstream.

Within the Conservative government of embattled Prime Minister Theresa May, Rees-Mogg is leading the charge towards a no deal brexit, an outcome desired by Russia due to its destabilisation of both the EU and NATO.

https://www.byline.com/column/67/article/2049

Is this claim true?

How reliable is this analysis?

— Added in response to comments:

How does the level of social media support (if any) given from Russian state-controlled sources to Rees-Mogg and other Brexiteers compare to the level of support they give to his political opponents? And to other UK political figures and viewpoints? Is this something that can be reliably quantified?

  • What is the claim? That some of the people who support him on social media are Kremlin bots? Or that the majority is? – user43646 Jul 24 '18 at 19:37
  • It would be strange if he wasn't They are known to support people from all sides of the political spectrum. Their goal is to create conflict and dissent, rather than supporting any particular ideology. – dont_shog_me_bro Jul 26 '18 at 9:31
  • @dont_shog_me_bro I’d find an answer which goes into detail about that - and quantifies the support they give to different sides of the UK political spectrum, if possible - very interesting. I agree the context is important. I’ll add to my question. – A E Jul 26 '18 at 13:01
4

The article reports that it is based on "Analysis conducted using open source tools at the University of Michigan known as the OSOME Lab or Truthy."

OSoMe does not offer any service which claims to track "Kremlin trolls". It does, however, have a service which detects "bots." The website for OSoMe's "Botometer" says:

Botometer (formerly BotOrNot) checks the activity of a Twitter account and gives it a score based on how likely the account is to be a bot. Higher scores are more bot-like.

Note the following: [...]

  • Botometer often categorizes "organizational accounts", like @BarackObama, as bot accounts.

Such approaches are known to be not free of false positives. Independent users of OSoMe report some glitches:

Algorithms and other automated tools are not useful in the effective identification of bots, and we should not base our conclusions solely on the results they provide. For example, the algorithm we used to detect bots – the “Botometer” developed by the Observatory on Social Media Project at Indiana University – did not take into consideration the linguistic aspect of tweets in Polish because it was trained on content created in English. It also flagged as bots many professional accounts that tweet frequently (news outlets) or are managed by more than one person (official accounts of politicians and political parties). This means that to effectively detect bots we need to dig deeper and examine the content of tweets, which not only requires a lot of time and effort, but also does not guarantee that it will be possible to say with certainty that a given account is a bot or not.

Karolina Iwanska and Anna Obem, "Digital Propaganda or ‘Normal’ Political Polarization?" Transitions Online, May 2018

Since OSoMe does not claim to track "Kremlin trolls" the real basis of the article's claim is currently unclear. The author cites specific Twitter accounts that he claims are Kremlin-run, but he offers no evidence for this, and in the past such claims have been made baselessly.

  • 7
    I didn't find this convincing. (1) Even if the tool has a very high false positive rate on individual accounts, we might trust the collective summary. (2) The connection to Russia wasn't made by detecting the bots were Russian, but by detecting that the accounts had "ties" to "known" Russian-backed sources. I remain skeptical about their methods, but I don't think the tool's limitations demonstrate the analysis is faulty. – Oddthinking Jul 25 '18 at 3:47
  • @Oddthinking and how'd you detect those "ties to known Russian backed sources"? All such claims I've seen are bogus, come down to "this account is positive about this right wing source, THEREFORE it's a Russian bot". – jwenting Jul 25 '18 at 11:52
  • @jwenting: Hence, I put "known" in quotes, and said I remain skeptical about their methods. – Oddthinking Jul 25 '18 at 13:27
  • 5
    To clarify: Premise #1: Authors had a gold standard basis for knowing some accounts were Russian-backed and some weren't. Premise #2: Authors could use the OSoMe API to develop a classification model that could discriminate between known Russian-backed bots and known not Russian-backed bots. They could then apply this to unknown accounts to determine whether they were (likely) Russian-backed bots. I am NOT saying Premise #1 nor Premise #2 is correct, but if they are both correct then this answer misses the mark. – Oddthinking Jul 25 '18 at 13:32

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .