The U.S. Senate on Twitter: Week One

Over the last five years, the social media platform Twitter has become a standard part of the communications package of U.S. senators.  An analysis of Twitter activity by senators in the first four days of the 115th Congress (Tuesday January 3 to Thursday January 6) reveals a large amount of communication with a great deal of variety between members of the Senate. During this period, the 100 members of the Senate posted out 1,792 Twitter posts (“Tweets”).  Many of these posts were accomplished impersonally (as with Senators’ speeches, statements and letters) through work delegated to hired communications staff.

The distribution of these Tweets is uneven. The office of New York Senator Charles Schumer posted the largest number of Tweets during the four days at 79, with Texas Senator John Cornyn not far behind at 69 Tweets. These two most voluminous Twitter users directed their posts in different ways: three out of five of Sen. Cornyn’s Tweets mentioned or replied to another Twitter user, while Sen. Schumer broadcasted his Tweets slightly more than half the time without any reference to any other Twitter user.  While both senators had much to say, Sen. Schumer acted as more of a broadcaster and Sen. Cornyn acted as more of a communicator.  The least communicative Senator on Twitter was Thad Cochran of Mississippi, who only posted one Tweet during the first four days of the 115th Congress:

Thad Cochran's one and only Twitter post during the first four days of the 115th Congress was directed toward Vice President-Elect Mike Pence

For a member of the U.S. Senate, Sen. Cochran’s single Tweet went relatively unnoticed, with only 5 retweets, 24 likes, and 14 replies. Vice President-Elect Mike Pence did not respond to Sen. Cochran’s outreach.

Those senators who do not communicate tend not to be the recipient of communication. Sen. Cochran, for instance, was not mentioned by any other senator during the new Senate’s first week.  Sens. Schumer and Cornyn, on the other hand, received multiple mentions from other senators during the period.  The most mentioned senator during the first four days of the 115th Congress was Catherine Cortez Masto, the new Senator for Nevada.  Most of these mentions by other senators were messages of welcome, although some noted her work, as in this retweeting message from Sen. Schumer regarding the new Senate’s plans to dismantle the existing health care structure:

Senator Chuck Schumer Retweets Senator Catherine Cortez Masto on the repeal of Health Care for millions of Americans

Patterns of communication between members of the Senate via Twitter tended to be partisan, as the following social network graph of mentions and replies indicates. This network graph uses a “spring embedded” visualization technique so that ties (indicated via curved lines) draw connected nodes closer to one another:

Twitter network of United States Senators. Lines indicate mentions or replies. January 3-6, 2017. Red nodes are Republicans, Blue are Democrats, Green are Independents, and gray are non-senate accounts.

Most Democratic senators’ accounts tend to cluster close to one another (although Senator Michael Bennet of Colorado is the network’s only “isolate,” not mentioning or referring to any other Twitter account during the period), and most Republican senators’ accounts also tend to cluster close to one another as well. Interestingly, the two Independents of the Senate, Senators Angus King of Maine and Bernie Sanders of Vermont, are clustered closely to Democrats’ accounts, Sen. Sanders most markedly so. Sen. King clusters with Democrats in this period because he mentions the same non-Senate Twitter account that they do (as indicated in gray).

There are exceptions to strict partisanship. Many members of the Senate refer to the same non-Senate accounts across party lines, as in the case of Senator Heidi Heitkamp of North Dakota, represented as the blue dot in the upper left of the network graph. While Sen. Heitkamp does not directly converse on Twitter with Republican senators, neither does she converse with her Democratic colleagues.  Because Sen. Heitkamp mentions an account that is also mentioned by Senator Joni Ernst of Iowa, she is clustered with Republicans. Senator Tim Scott (represented as the red dot toward the bottom of the network graph) follows the same pattern, not directly mentioning Democratic senators’ accounts but mentioning a number of the same non-senate accounts that they do. Senator Joe Manchin takes this pattern of cross-partisanship through indirect contact to its fullest extent, referring to the Twitter accounts of Vice President Mike Pence as well as the news outlets Fox News and the news shows Fox & Friends and Morning Joe that are popular targets of communication for a number of Republican senators. In the network of Twitter communication, this places Sen. Manchin squarely in the midst of the Republican upper-half of the Senatorial network.

As the large number of gray-shaded accounts in the network indicate, members of the Senate spend considerable energy communicating to accounts outside the Senate. Some 471 accounts outside the Senate were targets of communication during the first 4 days of the 115th Congress. The most common target of communication during these days was President-Elect Donald Trump, who was mentioned in 17 senators’ Tweets. The next most common target was the account of Planned Parenthood, whose federal funding for poor women’s pap smears and contraception is under threat from Senate budget cutters.  Rounding out the top ten most-referred to Twitter accounts by senators are four Democratic senators, the collective account of Senate Democrats, one Trump cabinet pick (Governor Rick Perry of Texas), and two national news outlets.

Patterns of reference to particular media outlets are highlighted in the network graph below, which is identical to the graph above but which features the accounts of national news outlets with graphic icons. The three most commonly referred-to media outlets during the period were MSNBC (10 references), Fox News (8 references), and C-SPAN (7 references). The location of some of these news outlets is unsurprising. Right-leaning Fox News, Politico, and The Hill are referred to most commonly by by Republicans, and left-leaning NPR is referred to exclusively by Democratic senators. However, the centrist CNN is surprisingly only referred to by Republican senators, and the right-leaning Washington Times is referred by by both Republicans and Democrats in the Senate.

Twitter Network of the U.S. Senate from Jan 3-6 2017 with national media outlets highlighted as graphic icons

 

Data collection and visualization for this post was carried out with NodeXL software.

Track Social Networks… to Find the People Tracking You

As the course designer and instructor for an undergraduate social networks course at the University of Maine at Augusta, I am often asked why students should take the course. I think there are many answers to this question. One answer comes from a humanities standpoint: learning how to represent patterns in relationships with attention to meaningful visual cues can deepen understanding of design and lead to innovation in art. Culturally speaking, networks have geek appeal as sparkling and colorful objects lending panache to infographics. If critical thinking is important to you, you might be interested in network analysis for the challenge of mastering multidimensionality and matrix mathematics; as you work at network puzzles you’ll develop your logical and quantitative reasoning ability. But these appeal aren’t all: the study of social networks can be practically useful, too.

One practical use of social network analysis is highlighted by the Disconnect extension you can add to your Chrome, Firefox, Safari, or Opera internet browser…

worried faceI should break in here. Whenever you read "extension you can add to your internet browser," you should begin to get nervous. Many add-ins, add-ons, and add-arounds to your internet browsing or Facebook or Twitter experience are so colorful and fun to play with. But they have a second purpose lurking behind the colorful and fun one: to track your movement across websites so someone can sell data about where you go and what you do. But when consulting Disconnect's privacy policy, I was pleasantly surprised to discover that the Disconnect extension collects information about you only minimally and doesn't sell information to advertisers: "Disconnect never sells your personal info.... Our browser extensions don't collect any of your personal info. Unlike most websites, our site doesn’t collect your IP address."

… so as I was saying, the Disconnect extension available for most internet browsers makes use of social network analysis to share useful information about websites that let your data leak out to third parties:

If you install the Disconnect extension in your browser, then visit a website, it will create a network graph (or “sociogram”) with that website at the center, visually linked to other websites that are given data whenever you visit that site. By bringing those network graphs together for different websites, you can figure out how your personal information might be combined and how that combination might be harmful to you.

That might sound a little abstract, so let me make it concrete. Consider the mini-industry on the internet of “Print-On-Demand” apparel. On websites like CafePress, Zazzle and Skreened, you can browse through thousands of t-shirt designs made up by people like you. If you find a design you like, you can put it on a t-shirt that fits your style, order that shirt, and have it printed up and sent specially to you. The printer gets a cut of the profits, the designer gets a cut of the profits, and you get just the shirt you want.

While these print-on-demand services are offering you a service that makes them a little money, are they harvesting your data on the sly? To find out, I activated the Disconnect extension in my browser and visited the CafePress, Zazzle and Skreened websites. Disconnect produced three sociograms, which I combine to form the network graph you see below:

How the Skreened, CafePress and Zazzle websites track your visits: February 2014

The above image is current as of February 2014, and represents an change in tracking since the last time I looked at these websites in December of 2012:

Skreened, CafePress and Zazzle website tracking technology habits: December 2012

There are a number of patterns to notice. Consistently and by a wide margin, CafePress has been sending information about you to the largest number of third-party websites. Over time, on the other hand, Skreened and Zazzle (to a lesser extent) have started to catch up, sending more information about you to other companies. Those companies include Lucky Orange (“We don’t just tell you who is on your site, we show you what they are doing”), Monetate (“helping you understand your customers’ situations, behaviors and preferences”), Retention Science (“analyze & predict customer behaviors”), and Tell Apart (“If you’ve ever clicked on an ad for a pair of shoes that seem like they were made for you, Tell Apart may very well have been responsible“).

When the practices of individual websites such as CafePress, Skreened and Zazzle are combined into a network, we can find points of overlap. CafePress and Skreened send their information to three websites in common: doubleclick.net, google-analytics.com, and googleadservices.com. Each of these services tracks users by IP address, so that your behavior at CafePress and your behavior at Skreened can be combined: these data mining companies can bring together your behavior at CafePress and your behavior at Skreened to figure out aspects of your identity and preferences that might not be apparent if they had access to only one of the websites. All three websites send data to googleadservices.com, leading to even more detailed insights about you. Would you be surprised to find out that doubleclick.net also receives information about visitors from nytimes.com, foxnews.com and amazon.com? Would it surprise you to know that doubleclick.net is owned by Google, bringing this overlap into even sharper focus?

Looking at simple lists of the third-party recipients of your information on a website can give you a rough sense of how leaky an individual website is. Looking at the network overlap in recipients tells you which of those recipients are likely to be learning the most about you, constructing an increasingly accurate virtual you for sale.