Stanford Internet Observatory’s three takeaways for using CrowdTangle to find evidence of and study influence operations:
Identify Pages of interest and put them into Lists.
Download Historical Data CSVs to analyze posting frequency and keyword trends.
Use Intelligence to identify suspicious engagement behavior.
The Stanford Internet Observatory (SIO) published a white paper this fall using data found in CrowdTangle to identify and analyze Russia-linked influence operations in Africa. Researcher Shelby Grossman also digs into the findings on Twitter. We talked to Research Manager Renée DiResta and Research Scholar Shelby Grossman about how SIO used CrowdTangle to research influence operations.
What are you researching?
The SIO team is researching inauthentic information operations in many parts of the world, including Africa, East Asia, and Eastern Europe. For this project, we focused on inauthentic information campaigns originating from entities linked to Yevgeny Prigozhin, a Russian businessman with ties to Russian president Vladimir Putin who was previously indicted by the US Justice Department.
What have been the results of your work?
We identified a cluster of Facebook Pages targeting Libya tied to Prigozhin, and Facebook shared with us dozens of related Pages they had previously been investigating. These Pages targeted Madagascar, Central African Republic, Mozambique, Democratic Republic of the Congo, and Sudan. As a result of this collaboration, Facebook removed 35 Facebook accounts, 53 Pages, seven Groups and five Instagram accounts targeting these countries.
We found, first, that the strategies of the Pages varied widely across countries. In Mozambique, the Pages were created in September 2019 and existing to bolster the ruling party before and after elections in October. In Libya, some Pages bolstered a rogue general while others supported one of Muammar Gaddafi’s sons – both of whom are seen as potential future presidential candidates. Second, we observed that these Prigozhin-linked entities appear to be franchising out their disinformation operations. For example, the Libya Pages may have been managed by an Egyptian digital marketing firm. The Central African Republic Pages had administrators in Madagascar.
Image courtesy of the Stanford Internet Observatory
How have you used CrowdTangle?
We’ve used CrowdTangle in many ways. We think of CrowdTangle as being useful both for discovery (e.g. identifying influence operations), and for studying what is actually going on with those influence operations. We primarily used Intelligence and Historical Data, as well as Lists to track operations. We’re super big fans of the Historical Data feature! The Dashboard overall was really useful for looking at historical Posts from certain Pages. On Facebook, it can take forever to scroll to the bottom of a Page’s history, but in the Dashboard we could start the feed the time that the first post was created, saving us a lot of time.
“CrowdTangle let us use built-in visualizations and also gave us the raw data so that we could create our own. Many people on our team prefer to analyze the data themselves in R, and we could do that with CSV downloads from Intelligence and Historical Data.”
How did you use Search and Lists to identify and track Pages that were part of the influence operation?
After Facebook’s Threat Intelligence team shared Pages with us, we created Lists of those Pages based on the categories that Facebook provided. Once we had the Lists, we could search for keywords across all Pages within a List. This helped us identify where certain phrases appeared first, and allowed us to search for Posts about specific topics.
How did you use Historical Data to analyze trends in Posts?
To analyze Posts made by the Pages of interest, we downloaded Historical Data CSVs for the Lists covering the entire period of their existence. From there, we created several visualizations, including Post frequency for all Pages by country, and tables that showed the average engagement by Page. Additionally, we analyzed Post text available in the Historical Data CSV to find keyword trends across Posts made by the Pages of interest.
How did you use Intelligence to identify suspicious engagement patterns?
We looked at the influence operations Pages and Lists in Intelligence to identify suspicious engagement behavior. There were certain patterns that pointed to inauthentic engagement, and Intelligence visualizations made it easy for us to identify them quickly.
Frequency of posting across the 11 Libya Facebook Pages. Source: CrowdTangle
Image courtesy of the Stanford Internet Observatory
What’s next?
One ongoing project is monitoring Taiwanese social media content in advance of their upcoming elections. Another project involves analyzing public Facebook posts about a recent attack on Libya’s capital. CrowdTangle will be critical for both of these projects.