I want to start with a basic example of a network that may be a helpful jumping off point. I presume most of you have seen a subway map. If you have, you’ve seen and worked with a network before! A network seems like it should be confusing, but really it is just a way to show relationships between different “nodes” or pieces of information. One professor I had in undergrad used this analogy to introduce network theory and, interestingly enough, my professor at UNC used the subway analogy as well to demonstrate her point. Obviously, it works!

  • Figure 1. Part of the New York City Subway Map
  • Figure 2. Part of New York City Subway Map With Adjusted Nodes

Let’s take a look at this image of the NYC subway map. Each station represents a node on the network, and each line in between the stations is an edge. Depending on your discipline, the terms node and edge may vary, but for clarity’s sake I’ll stick with these. An important concept that is related to nodes is their centrality. Centrality very simply measures the importance of a node. One way of thinking about this is how many connections are being made to a certain node, those with more connections are more “central.” Let’s look back at the subway map as an example of this concept. In Figure 2 I’ve adjusted some nodes; I’ve made two station’s nodes larger to represent that more lines converge there. Times Square in Manhattan and Atlantic Terminal in Brooklyn both are meeting points of multiple lines so I’ve enlarged their nodes to note their centrality. This makes the network graphic more useful. Without the adjustment, it seems as though the east 23rd street station on the 6 line and the Times Square station are the same size, which if you’ve ever been on the NYC subway you know is definitely not the case. Another way that changing nodes would be helpful in this case would be to show at which stops the local and express trains overlap. Let’s look at the two green lines, the 4/5 express trains and the local. They are depicted side by side and in places where all three stop there are simply 2 nodes. It could be easier, or at least a more simple graphic, if you had one green line representing all three train lines and smaller nodes would represent local stops by larger nodes would represent where both the local and express trains stop (since there are no stops that are only express and not local this would work). But enough about subways…

If you’re still not comfortable with the basics of networks, see Scott Weingart’s post on “Demystifying Networks.” Weingart does a great job of explaining the basic terminology, different types of networks (bimodal or multimodal for example), and how to read them. Something that Weingart mentions that I want to highlight is that a lot of context is stripped away in order to achieve readable network. I highlight this because the subway example doesn’t necessarily demonstrate the risks of reducing information well.

Let’s take a look at more examples

I want to preface this section with my usual cynicism. While I definitely think networks are useful, I think they are best used as presentational tools rather than research tools because of how much information is stripped from them. While they are helpful to include in presentations or to inspire research questions, applying research questions to them risks oversimplifying the issue. With this caveat in mind, let’s look at some of the examples that were brought up in class.

Let’s look briefly at how Scott Weingart introduced networks and an example he provided. In the example he provides a bimodal network. If we remember from earlier, this means there are 2 types of nodes, they can be either books (in blue) or authors (in red). In this example, he also has two types of edges. The black lines represent the relationship between author and book, i.e. who wrote what. The pink edge shows authors who collaborated with one and other. This is a relatively clear example of a network. There is not much overlap between concepts and relationships are obvious. It is almost clear to the point that you may ask: why bother with a network graphic? That’s exactly what I asked when I first read his post at least. The configuration of the different lines as swooping around and the apparently hierarchy of the nodes in the image makes it appear as if I should be getting more out of the image than I really need to. It seems just as easy that Weingart could have written “Edith Hamilton –> Mythology” on a horizontal plane rather than swooping the line dramatically to connect two floating bubbles. Again for me this represents an instance where it is definitely possible to make basic facts visual, but why. Obviously he merely intended this as a simplified example, but I think my point can be applied more broadly.

ORBIS: The Stanford Geospatial Network Model of the Roman World

Let’s talk about a finished example for a second. Take ORBIS: The Stanford Geospatial Network Model of the Roman World. This project is a massive undertaking. Poke around on it for a few minutes and you’ll realize not only how much labor went into making it, but how much data and research went into making it accurate. When I first interacted with the site I skipped the introduction and tutorial, and let me tell you that was a mistake. This is a fabulous tool, but there are so many moving parts it is difficult to use intuitively. At first glance it definitely doesn’t look like a network in the sense of the subway map or author/book examples I’ve discussed up until now. However, we can see that it is demonstrating a system of relationships in a visual way. Different places act as nodes and travel routes as edges that show the complex interaction of trade goods and distribution in the Roman World.

Screen shot of using ORBIS

I looked at the fastest route you could take with a porter (on roads) from Constantinopolis (in modern Turkey) to Corduba (in modern Spain). The fact that the menu on the left has so many ways to specify the travel is already amazing– you can specify you want to take an ox cart or a donkey rather than walking! We talked a lot in class about how all of us felt like we were using GoogleMaps when we were choosing the routes on the website. It was truly fun to play around with. But back to the point, this is still a network analysis! You can see the nodes on the map as the little circles that represent locations and the edges as the travel route. Thinking through networks in this way as not only a conceptually simple map, but also as something that is so malleable and manipulatable is really important. You can have the same base map (here of the Roman World) and demonstrate multiple networks on it using different nodes and edges in it. A cheaper route to Corduba from Constantinopolis may take me a completely different way via different roads (or edges) and stopping at different cities to restock (so at different nodes).

My own example

Network showing where some artists in the Tate collection were born, created using Palladio

Using that public Tate collection data I’ve used before, I created a network showing where artists were born. I used Palladio, which is an online tool. I found it quite easy to use. Honestly, once you upload or copy and paste your data into it it does pretty much all the work for you; really all you need to do is organize an Excel sheet with your data. Although the screenshot is small, you can see that there are nodes that denote artist names, and nodes that are locations of where they were born. I’ve sized the location nodes so that nodes with more centrality, which if we remember means more edges connect to it, will appear bigger. In this screenshot we see that London is a much larger node than Beijing because many more artists were born in London. I have a similar critique to Weingart’s example where the random placement of the nodes seems to imply something even though it really doesn’t. It doesn’t bother me as much in this example, but I still can’t figure out a rhyme or reason to the placement of the nodes on the graphic. It would be interesting to overlay this example onto a map perhaps, since it seems silly to have London, Venezuela, and Beijing right next to each other here when the network doesn’t seek to show any relationship between the places.

This network is simplistic and I’m not sure exactly how I’d use it in my art historical research, but I can see a multiplicity of uses for museum administration and acquisition meetings (sensing a theme anyone?). It is very easy to visually see which locations are represented more strongly in the collection. For a general public this could also be helpful as an introduction to the museum collections. I can envision this type of network being implemented into a larger interactive digital collection initiative like the Artlens Wall at the Cleveland Museum of Art which we’ve discussed in class.

I want to quickly mention we also played around with Gephi, but I wasn’t crazy about it. It did give you a few more options, but I didn’t think it was as intuitive to use as Palladio so at my level I’d likely stick to Palladio.