Thread by @shelbygrossman

1/ Excited to have a new paper, “In-House vs. Outsourced Trolls,” forthcoming in @polcommjournal with @noupside and @aasiegel drive.google.com/file/d/1CDSmkz…

2/ We theorize tradeoffs states face in running disinfo campaigns in-house v. outsourced. Outsourcing helps states tap digital marketing experts, saves $$ & provides plausible deniability. Running things in-house can have benefits from an op security perspective.

3/ We test implications of the theory by comparing two 2014-2018 disinfo campaigns on Syria that originated in Russia: the outsourced IRA campaign & the in-house GRU campaign. We look across Instagram, Facebook, & Twitter, leveraging data the platforms turned over to the Senate.

4/ Big picture: The GRU did what they always do, narrative laundering. Narrative laundering is more amenable to one-size-fits-all content. The outsourced IRA operation tailored content for particular audiences, and made the content highly partisan.

5/ Topic findings: The GRU posted about more one-size-fits-all content, like battlefield deaths in Syria. Meanwhile, the IRA posted on more partisan topics, eg Hillary Clinton and Obama’s Syria policies.

6/ And the IRA customized their posts for particular audiences, creating more partisan content, particularly on Facebook and Instagram, than the GRU. The figure below illustrates this for one topic, refugees.

7/ We find that the IRA used more clickbait/emotionally resonant language, again particularly on Facebook and Instagram, compared to the GRU.

8/ One of the paper’s unique contributions is the way we compare engagement. It’s not fair to compare direct Facebook engagement for the IRA & GRU. The GRU’s Syria operation was a narrative laundering op. They wanted their content to get re-posted elsewhere on the internet.

9/ The GRU’s Facebook operation primarily involved posting articles from the fake Inside Syria Media Center (ISMC) think tank. We searched sentences from the ISMC articles on Google (thanks @ShelbyAPerkins), and found 634 ISMC articles that were re-posted.

10/ We are making public the 1,541 URLs that were re-posts of the GRU ISMC articles: docs.google.com/spreadsheets/d… We have triple checked this list, but please let us know if you spot any errors. *We believe many of the websites that reposed GRU content did so unwittingly.*

11/ These are the domains that re-posted GRU Syria articles the most:

12/ For each URL re-post, we used CrowdTangle to assess how many times the URL was shared on Facebook, and how much engagement each share got. Even taking this engagement into account, the GRU’s operation still did not get as much engagement as the IRA’s.

13/ We hope this research spurs more research comparing in-house v. outsourced influence operations, particularly across platforms. We will be posting replication data shortly.