In previous chapters we studied social justice networks and affective publics from the bottom up. The interviews in Chapters 4 and 5 provided further insights into how social justice activists engage social media as part of their larger public presentation, as well as specific strategies such as hashtag hijacking. Now, we look at “gaming” the social media networks for political ends. Our case study is the 2016 presidential election in the U.S., where we visualize the structure of Twitter networks from political hashtags across the ideological spectrum (e.g., #MAGA, #ImWithHer, and #FeelTheBern). We look at network structures across the political spectrum and examine how different political strategies might lead to different network structures. Moreover, we look at the extent to which trolls and bots might influence these Twitter networks, as well as the impact that individual Twitter posters had during critical moments of the campaign, especially during the presidential debates between candidates Donald Trump and Hillary Clinton. A related purpose of this set of data visualizations of Twitter activity is to measure the relative popularity of tweets from Russian-based trolls and bots compared to those made by the candidates, celebrities, and other social media influencers. For instance, a Pew Research Center study predicted that it is bots that post about two-thirds of tweeted links to popular websites, rather than humans.1

This set of Twitter data visualizations is derived from an automated extraction process developed in Python 2.7, which allows us to search for specified terms in the Twitter historical archive, including such things as hashtags and words in tweets. The program stores these results, sorted by day, as a flat JSON file, formatted in two ways, with one data structure for exploring tweet-retweet relationships, and another data structure for viewing basic descriptive statistics about the search terms. Tweet-retweet relationship data structure builds arrays of nodes and links based on the Twitter historical archive. Nodes consist of users in the searched data, and links are built between those users and others who retweeted them. In this process, we preserve important data, such as tweet text and time of tweet, for more detailed exploration. These nodes and links are visualized using a modified D3.js force-directed graph that is filterable by time. We built a data visualization dashboard with the crossfilter.js and dc.js libraries for filtering and visualization; the descriptive statistics visualizations allow for interactive explorations of these data. Further, we use a dynamic/interactive version of these graphics, which are presented here. Figure 6.1 shows a static look at this graphic.

Figure 6.1

The pro-Trump continent. Pro-Trump hashtags are represented in blue.

From this visualization of the pro-Trump tweets, they show what we call a “continent effect”—a massive conglomeration of activity around pro-Trump hashtags (most notably #MAGA). Looking at the time period around the three presidential debates between candidates Trump and Clinton, pro-Trump tweets tend to amass in large continents on Twitter, as compared to pro-Clinton tweets that (at most) form small islands. Interestingly, however, the pro-Clinton tweets were actually greater in number than the pro-Trump tweets but were more widely dispersed. In some sense this is analogous to the election results, in which Clinton wins the popular vote but loses in the Electoral College based on how votes amass within individual states. While the number of pro-Trump tweets were fewer than the pro-Clinton ones, they were arguably a more impactful force, and are more noticeable in their conglomeration on Twitter. Compared to pro-Clinton tweets, pro-Trump tweets were more cohesive, as they created immense nodes (which, ironically, are in contrast to Clinton’s campaign message of “Stronger Together”). In terms of Twitter activity, pro-Clinton tweets were scattered apart. Comparing these data visualizations on Twitter to the story elements of the campaigns may show us something that traditional polling failed to predict in the election outcome. At the very least, Twitter visualizations are another way to make sense of the election that the polls did not.

Based on this initial analysis of the data, we wanted to find out which particular Twitter accounts were the most influential in creating pro-Trump continents. For instance, was Donald Trump’s personal Twitter account the most influential, or that of a particular celebrity, or those produced by Russian-based trolls (e.g., the fake “Jenna Abrams” account with 70,000+ followers that was created by the Internet Research Agency from Russia)? To discover this, we performed a “knockout experiment” in which we see the impact of individual accounts from the data visualizations by taking them out of the universe to see how it changes as a result. These “knockout” experiments borrow from genetic theory about what happens when you “knock out” genetic code within a DNA sequence to test the impact that a particular code piece has on the entire sequence. To our knowledge, this has not been done in social network analysis before, except on small scales and broad theories.

Additionally, we wanted to examine “broker figures,” which are Twitter accounts that do not necessarily provide new content but provide bridges in the social network through their retweets and number of followers. According to basic social networking theory,2 if you take out the bridges, you diminish the network. For example, three bot-generated tweets appear at the heart of a pro-Trump continent that feature Cyrillic script, as well as the #Ukraine hashtag. Other examples are references to the iPhone7 release in tweets linked to pro-Trump posts, thereby connecting unrelated tweets about the release of a popular mobile device to pro-Trump discourse (see Figure 6.2). It appears that bots were gaming the system (or hijacking hashtags) by utilizing some of the most popular (although unrelated) hashtags to increase the exposure of pro-Trump messages, which is a subtle and insidious form of influence. Beyond election messaging, this use of bots challenges the utopian view of social media and social networks, as large nodes can be artificially manufactured with insidious actors trying to game the system.

Figure 6.2

Broker figures and gaming the system

You can perform your own “knockout” experiments with this data set and observe the growth of several subnetworks, which appear to be right-wing influencers that are not part of the Trump campaign. In essence, the subnetworks of pro-Trump nodes have metastasized and become a highly resilient network overall. Even if you take out the center node, from the Trump campaign, the broader network is still there. Some of the most prominent of these subnetworks are nodes formed by the “bfrazer747” node (see Figure 6.3), the “TeamTrump” node (see Figure 6.4), and the “USA4Trump” node (see Figure 6.5).

Figure 6.3

bfraser747

Figure 6.4

TeamTrump

Figure 6.5

TeamTrump USA4TRUMP

Furthermore, even the Twitter accounts of well-established legacy news outlets do not have the same kind of scope of network. See, for instance, ABC’s node (Figure 6.6), CNN’s node (Figure 6.7), and MSNBC’s node (Figure 6.8).

Figure 6.6

ABC node

Figure 6.7

CNN node

Figure 6.8

MSNBC node

If you look at the network that surrounds Hillary Clinton’s account, the pro-Trump subnetworks have overwhelmed that landscape too. While Trump’s Twitter account has all of these subnetworks budding off of him, not all of them are leading back to him. Rather, the pro-Trump subnetworks appear to have their own center of gravity, and have even co-opted opposing hashtags, such as “#ImWithHer”. In summary, the pro-Trump dominance on Twitter is multi-tiered and even legacy media outlets are neutralized in their scope.

Gaming the 2016 Twittersphere: a closer look

Based on the research previously presented, we sought to explore how politically right-leaning entities appear to be more effective at using Twitter to promote their political inclinations and presidential candidate. The previous results (illustrated in Figures 6.1–6.8 above) might indicate that this is the result of troll activity, especially how pro-Trump Twitter accounts overwhelmed the node around Hillary Clinton. However, this is still only an assumption about the power of trolls and bots in monopolizing and manipulating the Twittersphere in support of candidate Trump during the 2016 election.

Our subsequent analysis presented below is based on data visualization methods that stemmed from questions about the influence of troll-like behavior from real users who are utilizing troll strategies to support their positions (e.g., inclusion of emojis alongside texts, the insertion of random plug-ins regarding unrelated topics, and the use of extreme and aggressive language).

Definition of terms

In this subsequent analysis we use the term “right-leaning” or “the right” to categorize Twitter handles supportive of candidate Trump, as well as handles expressing anti-Hillary Clinton, anti-Democratic Party, or anti-Obama statements. While many express pro-Trump sentiments directly by positively referencing the Twitter handle @realDonaldTrump, others express their support by lambasting Hillary Clinton while simultaneously using “left-leaning” hashtags such as #ImWithHer alongside the use of right-leaning hashtags such as #MAGA.

We use the terms “left-leaning” or “the left” to categorize handles with pro-Hillary Clinton content. More often “the left” is represented by handles strongly opposing Donald Trump.

Methodology

This subsequent study includes 1 percent of Twitter data from the date ranges of September 25–27, 2016, and November 6–8, 2016, and examines the data in three ways:

1. We examine the centrality of the major nodes in both degree and betweenness, as represented through numerical data and visual representations of these networks.
2. We look at the co-occurrence of the major hashtags during these same two time periods, where tweets from both the left and the right are analyzed to determine the use of hashtag co-occurrence by parties supportive of either side of the debate. The top 50 hashtags for betweenness and degree are used to determine the frequency of co-occurrence. Numerical graphs as well as bar graphs are used to dictate both the number of hashtags used by each group and the rate at which each utilizes co-occurrence within their network. Combined with the bar graphs, networks of the left and right, as well as the larger network, where the top eight hashtags are highlighted, provided visual representations of co-occurrence.
3. We performed a series of “knockout experiments” in which major nodes within the larger network are removed to determine their influence on the broader network. Visuals examine the larger network within September and November and highlight the major nodes essential to the structure and stability of the network. In addition to knockouts, this section includes visuals where major nodes are highlighted to present and compare their places within the network.

Centrality

When looking at the centrality of specific Twitter handles throughout September 25–27, 2016, we found that @realDonaldTrump (the official Twitter account of candidate Trump) was the highest in degree (0.0456), and @HillaryClinton (the official Twitter account of candidate Clinton) was the second highest in degree (0.0311) (see Figure 6.9):

As further seen in Figure 6.9, a majority of the handles that are highest in degree are pro-right handles (i.e., realDonaldTrump, DanScavino, LindaSuhler, bfraser747, magnifier661, TeamTrump, CarmineZozzara, The_Trump_Train, and StatesPoll). These major nodes hover around the two candidates as the first debate on September 26, 2016, unfolds. These major handles attract the most Twitter volume and draw the most attention to the debates surrounding their favored candidate, Donald Trump.

Donald Trump’s official Twitter account, @realDonaldTrump, also rates highest in betweenness (0.0990), while Hillary Clinton’s @HillaryClinton (0.0805) places second (see Figure 6.10).

As illustrated in Figure 6.10, the top handles in betweenness are mostly supporters of the right. Again, this visual showcases the strength of the right within the network and their role as an integral part in connecting others to the network. Some of the same handles highest in degree resurface as highest in betweenness. Again, most are those supportive of Trump and the right.

Also telling is the use of hashtags by the candidates during the September 25–27, 2016, time period around the first presidential debate. In Figure 6.11 we can see that the #MAGA hashtag is the most dominant. However, @HillaryClinton does not use any hashtags in her posts, except #LoveTrumpsHate, and only once during the days encompassing this first debate.

Figure 6.11

Visual of the September 25–27 network

In Figure 6.11, the major green node in the center representing @HillaryClinton remains the standard node color, green, meaning this node does not contain recognizable or popular hashtags; it is surrounded by various pro-right nodes. Therefore, we might infer that the @HillaryClinton handle maintains its high ranking (second within the network) due to the many right-leaning handles who use her handle as a mechanism to attract viewers to their own views.

However, @realDonaldTrump utilizes the #MAGA hashtag in several of his posts, as well as #MakeAmericaGreatAgain and #TrumpTrain. Many of the other pro-right handles employ these same hashtags in their own tweets—and alongside several left-leaning hashtags, which will be addressed later when examining co-occurrence.

Now let us look at the data and visualizations from November 6–8, 2016, which was the three-day period leading up to the election. Again, @realDonaldTrump rates highest in degree (0.0533), but @HillaryClinton drops down to third (0.0226). Moreover, Donald Trump, Jr. (@DonaldTrumpJr), candidate Trump’s eldest son, has moved into second place.

As seen in Figure 6.12, the majority of handles that are highest in degree are pro-right handles (@realDonaldTrump, @DonaldJTrumpJr, @DanScavino, @LouDobbs, @EricTrump, @LindaSuhler, and @TomiLahren). Similar to the previous period of September 25–27, the major nodes of activity hover around the two candidates, as the major handles attract the most twitter volume and draw the most attention. The most noticeable difference in the November 6–8 time period, though, is that the second highest in degree is another Trump supporter, @DonaldJTrumpJr, who had overtaken the other candidate in the race—@HillaryClinton—within the Twittersphere. While there is also a growth of some left-leaning celebrities, such as @ladygaga and @rihanna, they are seen in the periphery, rather than at the center of the action (see Figure 6.4 for more visual details).

Additionally, two of the handles in Figure 6.12 besides @realDonaldTrump, @DanScavino, and @LindaSuhler are present among the top handles for degree in both the September and November time periods. @HillaryClinton is the only repeated left-leaning handle in both periods. This may help to explain the strength of the right and their ability to create a more inclusive and stronger network. These handles could be constituted as “loyal followers” since they play a leading role in September, while also maintaining their place within the network for the month of November. This is not the case with the left, whose supportive handles play a less consistent role within the Twitterverse. Although the support for the left within this snapshot seems to become more equitable than it was for September (meaning there are more Twitter handles who are left-leaning among this list), the lack of consistency mentioned above may account for the strength of the right’s network, despite the fact that more left-leaning hashtags are coming out to support their candidate. Perhaps they came out too late, or simply did not conform to similar themes and messages to be effective?

As Figure 6.13 shows again, many of the handles are pro-right and vary in their hashtags (discussed later in the section on co-occurrence). Similarly, two of the pro-right handles (besides @realDonaldTrump), @DanScavino and @LindaSuhler, are present in the top handles for betweenness for the September and November periods; and again, @HillaryClinton is the only repeated left-leaning handle. Again, this is not the case with the left, whose supportive handles play a less consistent role within the Twitterverse.

Figures 6.14, 6.15, and 6.16 show how central Trump is to the network; however, the various nodes surrounding both candidates are right-leaning handles. In comparison, many on the outskirts are the top left-leaning supporters in terms of degree. In this visual (Figure 6.16), @HillaryClinton (the green node below @realDonaldTrump) is still central to the network but is virtually dwarfed by the density of clusters and individual nodes surrounding @realDonaldTrump. Additionally, when comparing these two major handles and their edges, it is clear that more edges stem from @realDonaldTrump compared to those stemming from @HillaryClinton. While @HillaryClinton contains several edges connected to nodes along the periphery, @realDonaldTrump is much more connected to other nodes, where their connections through distinct edges are prominent.

Figure 6.14

Visualization of the November 6–8 network

Figure 6.15

Visualization of the November 6–8 network

Figure 6.16

Visualization of the November 6–8 network

Co-occurrence of hashtags

We took the top 50 handles in betweenness from the time period September 25–27, 2016, to address the co-occurrence between these hashtags. Across the top and down the left side are the nine most popular hashtags (see Figure 6.17). Each box includes the number of times the hashtag along the left side is paired with the hashtag across the top. From left to right diagonally going down are the total number of hashtags within this network (see Figure 6.17).

The two most popular hashtags among the top 50 handles include #MAGA and #ImWithHer. #MAGA occurs in 26 of these top 50 handles, whereas #ImWithHer is present in just 5 handles. #NeverHillary also outnumbers #ImWithHer with 12 hashtags present among these top 50 handles as does #TrumpTrain with 7 hashtags and #TrumpPence16 with 8 hashtags. Co-occurrence among #MAGA is also much higher than #ImWithHer. While the use of the of particular hashtags in great number might be indicative of support, perhaps, more importantly, it inherently conveys power simply by its usage. Looking at the first column under #MAGA in comparison to the column for #ImWithHer (outlined in red, see Figure 6.18) it is clear that #MAGA as a hashtag co-occurs with other hashtags far more than does #ImWithHer. In fact, the only two hashtags that co-occur with #ImWithHer are both right-leaning hashtags (#MAGA and #TrumpTrain). The rest do not co-occur.

Figure 6.19 highlights the comparisons between co-occurrence from the left and the right. As the bar graph shows, Trump supporters, or those representing the right, are more likely to layer their hashtags. For instance, in Figure 6.19 the term MAGA ranks highest in terms of co-occurrence. Each color represents a hashtag paired with each one of the other hashtags. The amount of color visible among representations of each hashtag signifies the percentage of co-occurrence between these two hashtags. For instance, #MAGA is sixth in comparison to other right-supporting / Trump-supporting hashtags, such as #NeverHillary and #MakeAmericaGreatAgain; the left-supporting hashtags, #ImWithHer and #NeverTrump, experience very little co-occurrence. Additionally, #NeverTrump is accompanied by only Trump-supportive hashtags #MAGA and #TrumpPence16. Most notably, #MAGA is the most popular of the hashtags and is paired with the most other hashtags, including neutral ones like #Debate2016, and has a multiplier effect by using combinations of hashtags for one tweet.

Figure 6.19

Hashtag co-occurrence chart

Similar to Figure 6.17, Figure 6.20 identifies #MAGA as the prominent hashtag used among the top 50 handles in degree, among the 32 hashtags present in this instance. Again, the top used was #MAGA, followed by #NeverHillary (second), #TrumpTrain (third), #TrumpPence2016 (fourth), and then #MakeAmericaGreatAgain (fifth). All of these top handles are right-leaning and do a much better job of co-occurring with one another.

To measure highest in degree between September 26–27, 2016, we took the top 50 Twitter handles and looked at the co-occurrence between these hashtags. In Figure 6.21, across the top and down the left side are the nine hashtags, which are color-coded and note the highest in popularity. Each box includes the number of times the hashtag along the left side is paired with the hashtag across the top. The #MAGA hashtag dominates, towering over the rest with 32 uses; it is frequently paired with #NeverHillary, and again has a multiplier effect by using combinations of hashtags for just one tweet. Additionally, the neutral hashtag #Debate2016 is paired with #MAGA five out of six times, whereas it is only paired with the pro-Hillary hashtag, #ImWithHer, one of six times. The right is much better at grouping using neutral hashtags.

Figure 6.22 illustrates the same information in Figure 6.21 another way, but in more visual form. For instance, you can see here that the MAGA hashtag (a pro-Trump moniker) co-occurs with an array of other hashtags, including general hashtags (such as Debate2016), as well as other pro-Trump hashtags (e.g., AmericaFirst and TrumpTrain) and even oppositional hashtags, such as ImWithHer and NeverTrump.

Figure 6.22

Highest in degree

Additionally, the right has more overlap in its use of hashtags in comparison to the left, as the hashtags on the network only include the top in-betweenness and degree, meaning that there are fewer hashtags to compare between pro-left hashtag users.

In Figure 6.23, the left-leaning hashtags are represented as nodes of certain colors (colors representing hashtags) where matching colors hover around a major node. Two examples include the major yellow nodes, which consist of primarily of handles connected to these major nodes using these same major hashtags. Additionally, while there is some interconnection among the major nodes and the followers surrounding them within the left, overall they appear only sparsely connected. There are a lot of empty spaces within the network and weaker connections between each node.

Figure 6.23

All left-leaning popular hashtags, Sept. 25–27, 2016

However, when looking at the right-leaning hashtags (see Figure 6.24) a much more interconnected network is visible, as each node seems deeply intertwined with the communities of the other nodes. There is barely any “empty space” between these major nodes. Additionally, while the main color of this network is pink, representing “Election Day” (this is because of co-occurrence, where the #ElectionDay hashtag has replaced the hashtag before it), the colors surrounding each of the nodes consist of multiple colors representing the other hashtags. This is very different from the left, since the hashtags are much more diverse around the major nodes. Rather than only attracting handles or followers using the same hashtag, these major nodes on the right are able to attract those using various hashtags. This also reflects the success of co-occurrence used by top handles, since the major nodes are attracting followers that use various hashtags within their post as opposed to followers using the same hashtags.

Figure 6.24

All right-leaning popular hashtags, Sept. 25–27, 2016

When comparing Figures 6.23 and 6.24, it seems as if the color combinations flip. While the left is more diverse in the colors or hashtags that occupy its major nodes, the right consists of large pink nodes (representing the #ElectionDay hashtag), as many of the hashtags overlap and co-occur with other hashtags. The opposite is the case for the left, where the different hashtags/colors are seen among the major nodes. This indicates that less of the major nodes and their hashtags are co-occurring, since we don’t see a lot of the same color in a majority of this node. Additionally, the colors or hashtags surrounding these major nodes seem to differ markedly between the left and right. The left has many hashtags with similar colors surrounding the major nodes, which indicates that there is less communication between major nodes or communities where different hashtags are paired. In contrast, there is much more communication occurring between the right nodes, where multiple hashtags surround each of the major nodes. Again, this an indication that both the handles following these major nodes and the major nodes themselves are successful in pairing hashtags and therefore attracting followers using multiple hashtags, sometimes even many at the same time.

Knockout experiments

When looking at Figures 6.25 and 6.26, there appears to be little change when the @realDonaldTrump handle is removed from the network. Outside of the individual nodes once surrounding the @realDonaldTrump handle, many of the major connections appear to remain within the network. The edge networks also appear to remain intact, as the less significant handles and the edge networks are connected to other major nodes within the network.

Figure 6.25

Normal network, September 25–27, 2016

Figure 6.26

Network without @realDonaldTrump, September 25–27, 2016

Similar to the removal of @realDonaldTrump (Figure 6.26) the removal of @HillaryClinton (as illustrated in Figure 6.27) does little to disturb the network. While those hovering around this former node have dispersed, their connections to other major nodes have allowed these less significant nodes to maintain their space within the overall network. Furthermore, the rest of the network remains unscathed by the removal of @HillaryClinton from the network. While this may seem to suggest that the @HillaryClinton Twitter handle is unimportant within the network, @realDonaldTrump’s equally lackluster impact on the network after its removal suggests the importance of other major nodes/handles within the network and their ability to sustain support for their prospective candidates despite their candidates’ absence from the network.

Figure 6.27

Network without @HillaryClinton, September 25–27, 2016

Even without several of the major nodes (see Figure 6.28), the right’s network remains visibly intact. While the candidates’ Twitter accounts, @realDonaldTrump and @HillaryClinton, have some influence on the network, they are not the only major nodes keeping the network together. This suggests that strong political networks are made up of several independent nodes that support their preferred candidate, but also build a fan base (or community) around themselves. These make for a stronger network, where the major node, @realDonaldTrump, is not necessarily essential for the stability of the network.

Figure 6.28

Network without @realDonaldTrump and @HillaryClinton, November 6–8, 2016

When looking at Figure 6.29, many of the major nodes on the right (@EricTrump, @TeamTrump, @LouDobbs, @LindaSuhler, @rudygiulianiGOP, @WDFx2EU8) are removed along with @realDonaldTrump. Here the overall network is visibly starting to collapse. Nonetheless, even with @realDonaldTrump gone and with many of his top followers on the right removed, the network is able to exist and still contains many Trump-supportive hashtags.

Figure 6.29

Network without other major nodes, November 6–8, 2016

Figure 6.30 shows the major Twitter nodes supporting the right among the top 25 hashtags in degree. These nodes are highest in degree in the larger network, as well as in their connection to other parts of the network. Highlighted hashtags include only those supportive of the right. These nodes, or Twitter handles, include @DonaldJTrumpJr, @DanScavino, @LouDobbs, @EricTrump, @LindaSuhler, @TomiLahren, @TeamTrump, @rudygiulianiGOP, @Lrihendry, @WDFx2EU8 (which may be a handle belonging to a bot, or troll), @WeNeedTrump, @ChristiChat, @bfraser747, @mike_pence, @Stonewall_77, and @LaraLeaTrump. In comparison to the left-leaning handles, these nodes form a clear circle around the two major candidates (see Figure 6.30) and can be seen actively participating in connections not only between these major candidates and other supportive nodes, but also interacting with and connecting to one another, helping to form a circle, or rather star-like shape, encircling the candidates’ nodes (@realDonaldTrump and @HillaryClinton).

Figure 6.30

Right-leaning Twitter handles highlighted among the top 25 handles in degree for November 6–8, 2016

Comparing this to the major left-leaning nodes highest in degree (see Figure 6.31), the right-leaning handles within the network almost resemble an attacking force surrounding the two major candidates. It also seems like some of the major left-leaning supporters (again, mostly celebrities) fall outside the periphery of the graph, as demonstrated and discussed below.

Figure 6.31

Left-leaning Twitter handles highlighted among the top 25 handles in degree for November 6–8, 2016

In Figure 6.31 above, the following handles are highlighted: @rihanna, @ladygaga, @NormaniKordei, @ChrisEvans, @ddlovato, @JLo, and @thatbloodyMikey. It is clear that many of these nodes outside of @ladygaga (in the upper left corner) and @ddlovato (the one highlighted edge from the upper righthand corner) do not have a strong connection to the network. Many of these left handles consist of celebrities and are found along the periphery. In fact, outside of the two handles mentioned, many of these handles cannot be seen. It also seems as if they are attached more strongly and centrally to @realDonaldTrump, rather than @HillaryClinton, despite the fact that they are more supportive of candidate Hillary Clinton and use hashtags supportive of her. This indicates a lack of bridging and a weaker (or more dispersed) network in spite of the number of individual supporting Clinton on the overall network.

In comparison to the right-leaning hashtags, these nodes are far removed from the conversation, figuratively and literally. There is not only disconnect from the main participants in the network, but there also appears to be little conversation occurring between these major nodes. This speaks to the lack of left-leaning nodes among the top 25, but also addresses the inability of these nodes to speak to the larger issues and more popular hashtags promulgated throughout the network.

Major Findings

From the data presented here, the right is far better at gaining attention and traction within the network via hashtags, as demonstrated in the visuals and numerical value attributed to the betweenness and centrality of right-leaning hashtags. Meanwhile, bots and major news outlets are not prominent within the network, as the majority of the top-ranking handles (as dictated by betweenness and degree) are prominent Twitter users, such as politicians, or others with ties to the presidential race, including celebrities (especially during the November 6–8, 2016, range).

The right does a much better job of pairing hashtags alongside one another within their tweets. This is a product of the right using more hashtags, and also because the prominent hashtags in the network consist predominantly of right-leaning hashtags. Additionally, the right in several cases links left-leaning hashtags, such as #ImWithHer alongside #MAGA, in ways that promote their candidate of choice, while deriding his opponent.

While the handle @realDonaldTrump is a major figure within the network, the figures presented here, as well as the knockout experiments, show the strength of the right within the network even when their protagonist, Trump, is completely absent from the network. This is indicative of the strength of the right’s Twitter network. In fact, there was no major node that, when removed, was able to dismantle the network alone. It would take the removal of several of these major nodes for any distinguishable changes to occur.

That said, given the prominence of fake news that employed bots during the election campaign cycle of 2016, we might question what impact they had. While the right demonstrates a strong Twitter network in and of itself, we now consider the more difficult-to-measure impact of fake news, bots, and doublespeak on the Twitterverse throughout 2016, and lessons social justice movements might take from the right in their use of social networking applications.

However, a limitation of our analyses of the Twitter data sets in this section was our focus on hashtags instead of specific conversations. As we have noted in previous chapters, tweets that contain powerful images or resonant emotions can go viral without the use of hashtags. To be clear, we are not suggesting here that right-leaning networks are more cohesive based on the resonance of particular messages; rather, our analyses do show how virality can be manipulated through the use of hashtags. There is of course real power in other kinds of tweets (without exploiting hashtags) that have power in their own right and may ultimately affect social discourse on social media platforms. In the next chapter we take a deeper look at the political and commercial manipulation of social media discourse, as well as its input on new, information, and technology.