Welcome to the fourth course lecture for COM/SOC 375: Social Networks. In this lecture, we’ll consider the following subjects in a combination of text, images and video:
- Revisiting Network Centrality
- Networks are Everywhere: Maps as Networks
- Installing and Generating Networks with NodeXL
- Guidelines for Visualizing Networks
To prepare for this lecture you should read Chapters 4 through 6 of Analyzing Social Media Networks with NodeXL by Derek L. Hansen, Ben Shneiderman and Marc A. Smith. Be prepared to start work on Homework #4, which means installing NodeXL on your computer (or using it on a University student lab computer), generating a sociogram that contains contextual information about your family network, and obtaining information about that network using the “Graph Metrics” command.
Revisiting Network Centrality
A Centrality Cheat Sheet
In all of the comments posted to last week’s lecture on network characteristics, the most common theme had to do with confusion about measures of centrality in networks. In response, I offered to put together a quick network centrality “cheat sheet” that you could print out and use as a primer. The sheet is placed below — and just to clarify, it’s not cheating for you to use it! I’ve placed it here using a smaller “thumbnail” image. If you’d like a larger version to print out, just click on the thumbnail image to load a larger version:
Thinking Some More About Betweenness Centrality
Of all the uncertainty about centrality students have expressed, the greatest uncertainty seems to regard betweenness centrality. Although the calculation of betweenness centrality is fairly simple in the one-sheet primer I’ve placed above, it can become more complicated when pairs of nodes have more than one geodesic path between them. Let’s try thinking about betweenness centrality in a more detailed with the following video that goes through two examples, then asks you to try your hand at working out a third example.
How’d you do by the end?
Networks are Everywhere: Maps as Networks
Although matrices and sociograms (also known as “network graphs”) have been used by academics to characterize patterns of social ties for some eighty years, they didn’t burst into popular consciousness as “networks” until the last decade or so. Still, once you begin to think of the world in terms of social networks, network imagery seems to appear everywhere, and many of those images predate the popularity of networks.
You’ll recall from assigned readings in Derek Hansen et al’s Analyzing Social Media Networks and Lecture 2 that centuries ago Leonhard Euler came upon graph theory as a way of simplifying his daily walks through the city of Königsberg:
But the simplification of our travels through the physical world did not begin with Euler. Consider this detail from the “Tabula Peutingeriana,” a 1265 image of the roads of the Roman empire:
This section of the image, which shows modes of travel for Constantinople, Crete and Asia Minor, is not precisely a geographic map. Accurate depiction of waterways and landmasses have been sacrificed as secondary in order to highlight the information of primary importance: communities (which you might think of as the nodes of this network) and straight-line roads (which are ties connecting the communities nodes). The monk who drew the Tabula Peutingeriana wouldn’t have used the phrase “social network graph” to describe his image, but he was clearly thinking in a fashion parallel to that of modern network analysis.
In modern times, road atlases like the one you see below pay more attention to the precise location of some features — parks, highways, communities and bodies of water — in physical space. There’s even an inclusion of a scale key so that the physical distance between and along these features can be measured in miles. But even here, you’ll notice that complete physical accuracy has been sacrificed in order to highlight relational information. The actual appearance of Idaho and surrounding areas is obscured: mountains and plains have been painted over in blocks of white, or tan, or brown. Smaller communities and smaller roads are not included on the map. The information communicated here pertains the largest communities (our “nodes”) and the busiest roads (the strongest ties) are indicated, so that those traveling from one region of Idaho can take the shortest drive (or network “walk“) from one place to another.
A modern transportation map fully corresponding to a network sociogram is the following map of the “London Underground,” the subway system beneath the capitol city of the United Kingdom. Once a traveller heads underground, they can no longer see geographic features, and as passengers rather than drivers of the Underground cars the twists and turns taken by the cars are irrelevant; only origins, destinations, and connecting stations need to be known in order for people to get from place to place. Want to be communicated across the Underground’s transportation network? Just follow the lines between the nodes. For an equivalent network map in the United States, check out this graphic for the DC Metro.
What’s really interesting is that abstract maps like interstate atlases are confirmed through measurement of our collective behavior. Consider the Vulcan Project of Purdue and Arizona State Universities, which tracks carbon dioxide emissions across the United States. In the video below, Vulcan Project visualizers have directly mapped CO2 emission, indirectly mapping population centers and transportation routes that produce carbon dioxide. This CO2 map is a social network map of nodes and ties between the nodes.
Taking the network insight regarding maps down a level in scale from the inter-city level to the within-city level, the City Form Lab at the Massachusetts Institute of Technology has released a software tool for urban network analysis in physical space, to be used with the (unfortunately not free and open source) ArcGIS software suite. I don’t expect you to use this software, but I do want you to learn about it and see what kind of practical applications are possible for analyzing social networks across physical space. The skills to which you’re being introduced in this class can be instrumental in a professional career, as you’ll see. Please watch this short video:
This week’s homework asks you to install NodeXL, to move your family social network data from UCINET to NodeXL, and to visualize your network as a sociogram in a meaningful way. The following video walks you through the process of installing NodeXL:
And here’s another video that shows you how to create a network by entering edge lists in NodeXL, and also how to “visualize” a network graph in ways that show important aspects of your network’s structure:
Working with software like NodeXL can sometimes lead to surprises. A piece of advice: start your homework early so that you can have time to deal with unexpected problems. If you encounter any problems you can’t overcome with a little trial and error, be sure to get let me know.
As we continue our semester, the folks behind NodeXL at the Social Media Research Foundation have come up with a bit of a surprise, which you can see in a new “splash screen” that appears as you start NodeXL:
Some time soon — I’m not sure about the exact date — NodeXL will be splitting itself into two versions, NodeXL Basic and NodeXL Pro. This change will be a bit unsettling, but please don’t worry about the change. We’ll make adjustments to the course as necessary, depending on when the change occurs. I won’t ask you to spend money on the upgrade; that would be unfair, given that we didn’t see this change coming at the beginning of the semester. Look for a further update on this subject in the weeks to come.
Guidelines for visualizing networks
In a series of publications spanning his career, graph theorist Kozo Sugiyama attempted to develop a series of rules (and later in his life a series of rules for making rules) about how to create network graphs that are easy to read and interpret. See the “References” section below for some citations of books and articles in which Sugiyama discusses his reasoning behind these rules, a few of which are:
1. Lines should be straight.
2. Lines should be far apart from one another.
3. Lines should not cross or touch.
4. Lines should be easy to follow from one node to another.
5. Nodes that connect should be close.
6. Nodes that are most central should be close to the center of the graph.
7. Nodes that are similar in some way should be placed nearby one another.
You may want to think about some of Sugiyama’s rules as you complete your homework this week. These rules are not laws of physics, but rather principles of design that embody some assumptions about what network graphs are supposed to represent. As our ideas about what images represent change, the rules of network design should be expected to change as well.
Later in the course, we’ll conduct some analysis of online networks. As a preview of what’s possible with NodeXL software, you may want to follow along with this video example, which pulls a complete network of Twitter data from the Internet and reviews a few ways to organize the resulting network graph. Although Kozo Sugiyama’s rules seemed like a straight-ahead way to keep network graphs clear in the dawning years of the Internet Age, we’ll see that as the years have progressed and online interaction has become more intense, it has become important to bend those rules a little bit.
Now it’s time for you to get participatory again! For your in-course lecture participation this week, head out onto the internet (Google Images at http://images.google.com is a good place to start) and find an image of a network graph from somewhere out there in the Internet. Pin your image onto the Padlet below by clicking the appropriate “add an attachment” icon…
… and entering the URL (web address) of that image. Then review the image in terms of the quality of the visualization: does the visualization follow Kozo Sugiyama’s guidelines? Regardless of whether it follows Sugiyama’s rules, do you think the network graph is a good/sensible/useful visualization or a poor/confusing/useless one? Why is that?
(It’s best to use the Padlet, because the format lets us learn from one another. But if you can’t get the Padlet to work, please send your contribution to me by e-mail at email@example.com.)
Freeman, Linton C. 1977. “A Set of Measures of Centrality Based on Betweenness.” Sociometry 40(1): 35-41.
Friedkin, Noah. 1983. “Horizons of Observability and Limits of Informal Control in Organizations.” Social Forces 62(1): 54-77.
Sugiyama, Kozo, Shojiro Tagawa and Mitsuhiko Toda. 1981. “Methods for visual understanding of hierarchical system structures.” IEEE Trans. on Systems, Man, and Cybernetics SMC-11(2): 109-125.
Sugiyama, Kozo. 1987. “A cognitive approach for graph drawing.” Cybernetics and Systems 18(6): 447-488.
Sugiyama, Kozo. 2002. Graph Drawing and Applications for Software and Knowledge Engineers. Singapore: World Publishing Company Inc.