#IranVotes: A data analysis of Iranian Twitter during the 2016 parliamentary elections
Although Twitter remains blocked by the Iranian authorities, the widespread use of circumvention tools by Iranian citizens has allowed them to make use of it as a free and open space for public engagement around contentious and divisive political and social issues. Using a mixed-methods approach combining social network analysis with qualitative content analysis of election-related content of the Iranian Twittersphere during the 2016 parliamentary elections, researchers from Small Media and the Berkman Klein Center for Internet & Society were able to identify and analyze 46 clusters of users ranging from human rights activists through to reformist and conservative political commentators, technology advocates, and literature enthusiasts. In addition to these interest-bound clusters, they also observed that the network is home to extensive networks of everyday users, who share jokes, idle chatter, and flirtatious messages.
To begin, the researchers generated the following social network map of Persian-language Twitter accounts:
Image: A visualization of the Iranian Twittersphere.
The network structure was visualized using a physics model layout algorithm, overlaid with colors representing each account’s assignment to a group based on a clustering of network relationships. The resulting network map and structures that emerge reflect the individual decisions of Twitter users to follow other users, and the work of the algorithm that transforms these follow relationships into clusters.
Breaking down this network map further, 46 clusters were identified and classed into the six groups above based on the types of discussions and user engagements that took place. For example, clusters and conversations in the Cultural and Social Issues group looked like this:
Image: Group, cluster, and conversation descriptions.
Using these clusters, the researchers measured density and mutual tendency to gain a deeper understanding of each community.
Cluster density measures the level of connectivity among nodes in a particular network, and can be calculated as the number of connections between users in a cluster as a proportion of the total possible number of connections, while the measurement of mutuality tendency provides evidence for the levels of two-way information exchange and user engagement within a community.
From this analysis, it was clear that clusters within the Cultural and Social Issues group had the closest network connections:
Image: Clusters with the top mutuality tendency.
The censorship context
Gauging the locations of users in the network was a complicated task, and owing to many users’ reticence about listing their location publicly. 40.3% of users were identifiable by the time zone listed in the dataset. For those listing the ‘Tehran’ time zone, the team applied the tag ‘Inside Iran,’ and for those outside, they applied the tag ‘Outside Iran.’ Then, looking at user-specified location data, which was provided by 48% of the remaining untagged users in the network, they manually tagged these users based on the information they provided.
After completing this step, the team was left with 22.6% of users who either did not provide any identifying location data, or provided ambiguous information such as ‘Somewhere,’ ‘Nowhere,’ or ‘Neverland.’ They found that 50.7% of users stated they lived inside Iran, and 26.7% stated they lived outside the country.
Image: Image: User location by group.
Surprisingly, data on external links inversed some notions about censorship context in Iran. Despite being blocked in Iran, YouTube was the most widely linked-to website in the network, with 1826 links. Notable by their absence are Iran’s domestically developed social media platforms, such as Facenama or Afsaran, Instagram analogue Lenzor, or the Iranian version of YouTube, named Aparat. Although the Iranian government has a long track record of filtering internationally developed SNSs in order to push users towards Iran-hosted (and thus more easily surveilled) alternatives, such efforts appear to have made little headway within this network.
Image: Top information source links.
Differences between groups during the election campaign
In the absence of a rigid party system in Iran, candidates for legislative elections often run on electoral lists (or slates)—groups of candidates that align themselves with each other in an effort to boost their collective chances of electoral success. One such list—deemed the ‘List of Hope’ (امید_لیست (by former reformist President Mohammad Khatami—was at the core of reformists’ and moderates’ political hopes in the 2016 legislative elections. Grouping together the last remaining reformist candidates who survived the Guardian Council’s pre-election electoral purge with moderates and ‘least-worst’ conservatives, the List of Hope was backed to deliver victory to as many non-hardliners as possible.
On Twitter, the ‘List of Hope’ was mentioned numerous times within the core of the network, and was shared by a number of key opinion shapers including BBC Persian, Cafe, and Negar Mortazavi. In addition, it spread throughout the Cultural and Social Issues and Mixed Users groups.
Image: Users mentioning ‘List of Hope’, February 18-29, 2016.
Although reformists and moderates made effective use of online campaigns such as the ‘List of Hope’, they weren’t the only ones making use of SNSs to advance their electoral prospects. Conservatives waged a reactive campaign in opposition to the ‘List of Hope’ that attempted to characterize it as the ‘English List’—an electoral slate cooked up by the British government and the BBC to undermine Iran’s national independence and autonomy.
These accusations quickly made their way to social media platforms, where they became a discussion topic for conservatives and incredulous supporters of the List of Hope. Whereas the ‘English List’ was fairly widely discussed across the network, the ‘No2UK’ campaign was effectively isolated to the conservative sectors of the Twittersphere from which it originated.
At the conclusion of the project, the team found that scale of Twitter activity amongst diaspora Iranians and more liberal segments of Iranian society had two major impacts upon the political makeup of the Twittersphere: firstly, a general politicization of the Twittersphere; and secondly, the squeezing out of politically divergent voices—especially from conservative factions, who appear to have found their home on alternative (unblocked) social networking sites. Although it does not necessarily hold a hugely politically diverse or representative chunk of Iranian netizens, the Iranian Twittersphere does function as an important bridge to connect the country’s vast diaspora networks to politically engaged, reformist-leaning citizens living inside Iran.
Read the full report here.