Thread by @conspirator0

We've been posting a fair amount about tweetdecking and related forms of artificial amplication of late. Can we find evidence that similar techniques have been utilized by the Internet Research Agency?

cc: @ZellaQuixote

https://twitter.com/conspirator0/status/1175577488027766785

@ZellaQuixote One telltale sign of tweetdecking is highly viral recycled/plagiarized tweets. As it turns out, (at least) 11 of the 20 most-retweeted banned IRA troll tweets are copies of older tweets. All came from fake left-wing accounts, and all achieved disproportionate retweet volume.

@ZellaQuixote Tweetdeckers earn unusually high retweet volume because all the accounts in a "deck" retweet one another in a coordinated fashion, which shows as a dense cluster on the accounts' retweet network. We found a potential cluster of this sort in the IRA network (circled, upper left.)

@ZellaQuixote Did the accounts in that cluster have other traits associated with tweetdecking, such as the use of cloned tweets? Answer: yes. The same tweets were used over and over, both across accounts and on the same accounts.

@ZellaQuixote These accounts also used custom apps to post nearly all of their tweets. Although there are 329 accounts in the cluster, only a few dozen are active at any given point in time.

@ZellaQuixote Obviously, none of this proves that the Internet Research Agency engaged in tweetdecking, per se. It does, however, suggest the presence of similar methods of artificial amplification were in play.

@ZellaQuixote (Internet Research Agency datasets available here - only English-language tweets were included in this experiment)
about.twitter.com/en_us/values/e…
github.com/fivethirtyeigh…