Earlier this month, Internet researcher Lawrence Alexander published a network analysis on thousands of seemingly automated Twitter accounts, pointing to extensive information manipulation attempts on the RuNet. After the analysis garnered massive attention from Russian media and social networks, Alexander now addresses some of the skepticism about the bot networks and their provenance.
My social network analysis of over 20,000 bots posting identical sets of pro-Kremlin statements on Twitter, apparently in reaction to recent news and events, attracted quite a bit of media attention and proved to be the fodder for some contentious discussion on the RuNet, so I'd now like to address some of the criticisms.
Some Russian media commentators have downplayed the pro-Kremlin manipulation of the Twittersphere, claiming the automated bots and human-run accounts are created merely for marketing or spamming purposes. Sputnik News, for example, takes this stance:
…the goal [of the trolling] appears to be not to convert public opinion, but rather search engine optimization (SEO): search engines such as Google crawl the web for keywords and links, which are then used to give weight to certain web pages when search terms are entered.
However, this fails to explain the trolls’ apparent bias towards pro-Kremlin sentiment, and offers no suggestions as to who exactly might gain from running the accounts besides a vague “political consultancy firm.” It also doesn't consider that a deliberate disinformation campaign might employ tactics similar to spammers, both for convenience and for added ‘plausible deniability.’
In my analysis, I found little evidence of any commercial purpose behind the bot networks. Most tweets didn't link to premium sites or share dubious-looking links. And because I was able to establish that few of the accounts were genuine (i.e., run by real humans), it didn't seem likely they were part of a paid followers scheme, which would follow plenty of real people and organisations.
Other critics have questioned how much impact the bot networks could have, considering they rarely interacted with others.
The answer, I believe, lies in Twitter's Search and Trending features. A good way to illustrate this is to look at the way bots have been shown to spread news stories in their tweets.
In March 2015, Keir Giles, director of Chatham House's Conflict Studies and Research Centre and a fellow Russia and Eurasia Programme, noted repeated sharing of this RT article. In it, several interviewees attempt to cast doubt on evidence that flight MH-17 was downed by a Buk missile.
— Keir Giles (@KeirGiles) March 11, 2015
I used NodeXL to collect a Twitter Search network sample for the headline seen in the tweets: “Could SU-25 fighter jet down a Boeing? Former pilots speak out on MH17 claims.”
The resulting graph (below) shows the accounts sharing the article as circular ‘nodes.’ Those that are closely interconnected by follows or mentions are grouped into color-coded clusters. (Their size is based on their relative connectedness with others in the network.)
Prominent in the center—for obvious reasons—are RT (Russia Today's) accounts, @rt_america and @rt_com. The strongly connected upper green cluster shows a number of pro-Russian tweeters following or interacting: for example, @gbabeuf and @novorossiyan. But the anonymous bots are found mostly in the outer ring, indicating their characteristic lack of interaction and lack of following real users.
Just because the bots are isolated, though, doesn't mean they lack impact. Their repeated use of key words, phrases, and headlines affects Twitter's search results and trending topics, so anyone looking for tweets about MH17, for example, would have seen multiple instances of RT's story, boosting its perceived popularity. (There is no evidence that RT themselves have anything to do with the bots).
These strategies are explicitly prohibited in Twitter's rules—though this hasn't stopped large-scale bot networks from encroaching on Twitter news narratives, potentially influencing trending topics. Unless social networking sites strengthen their protections against mass account registration, it seems likely these manipulation efforts will continue.