Tool Trouble: Network and Spatial Methods in Art History

This talk was presented at the 2016 College Art Association conference at the session “Digital Art History: New Projects, New Questions” in response to three papers given by former participants in George Mason University’s 2014 digital art history institute.

Daniela Sandler discussed her classroom use of Esri’s Story Maps, Claire L. Kovacs her research on expatriate networks in Degas’ Paris, and Cynthia B. Kristan-Graham her digital mapping project of Chichén Itzá.

Network graphs and spatial maps aren’t new to art historians. Maps, we know well, and graphs? Just look at Alfred Barr’s diagram from Cubism and Abstract Art. Barr constructed his diagram to illustrate. The visualization of his preexisting argument was the goal. But network analysis is more than graphing, spatial analysis more than mapping. For Claire and Cynthia, it was thinking through questions about network structures and spatial relationships — the process of arriving at a visual graph or map — that constituted their research. In Daniela’s case, it was also process, rather than product, that constituted the key pedagogical moment: through mapping, her students became more aware of the way that they interconnected evidence when making larger synthetic arguments.

I think we are all on board with the idea that computation can aid in producing new evidence and questions. In my response, I’d like to highlight the challenges of charting a course between our colloquial notions of what networks and spaces are in art history, and the frames that network analysis and GIS use to reason about them. How can we start to shift from thinking about “tools” to thinking about “methods”?

We might be used to thinking about networks in a ad-hoc way like Barr’s diagram. His network merges together many types of constituents and connections in a muddy Gesamtnetzwerk that looks impressive, but actually flattens the differences, for example between an individuals’ affiliation with national identity versus with a stylistic school. Computational network analysis demands more careful precision. Claire spoke to this challenge in her presentation, as she has evidence for both correspondence networks as well as co-exhibition networks that each connected the same groups of people in Paris.

Rather than constructing one master network, it can be more productive to compare parallel or overlapping networks with the same people, but different patterns of connections. Historical sociologists do this all the time. One of the best examples is John Padgett’s foundational work on the parallel networks of marriage, economic, and political connections in Medici Florence. Though each network had common members, the Medici constructed each to have very different structures. By comparing the unique structures of each layer, Padgett illuminated how the Medici strategically isolated major Florentine rivals from one another. One wonders how this comparative approach could lead to a richer picture of the Italian community in Degas’ Paris.

I would also challenge art historians to explore the full spectrum of questions that network analysis can speak to. We can go beyond individuals to compare entire networks, or different phases of a network as it changed over time. I research how structural incentives in Netherlandish artistic printmaking led to shifts in the way publishers, engravers, and designers networked in the sixteenth and seventeenth centuries.

My evidence are the tens of thousands of print impressions in the British Museum and the Rijksmuseum, each one an index of several professional relationships.

From these, it is possible to construct a potential representation of the production network that might have resulted in these prints we know today.

Rather than collapsing 200 years of history into one static graph, like so, I instead think about the network as it might have existed at different points in time — which nodes, or artists, were active at that time, and how were they linked via the prints made during that period? For each of these slices in time, we can track certain network-level metrics.

Network centralization could illuminate whether the medium of engraving incentivized more centralized, or more distributed network structures at different points in history.

I am also interested in group assortativity: to what extent did the collaborative clusters formed by artists align with, or depart from, national or regional identities?

By tracking these metrics over time, we can produce informative plots rather than beautiful-but-illegible hairballs. The results suggest a printmaking history where seemingly major changes can be understood as the emergent result of a handful of deceptively simple and continuous incentives. This view presents a counterpoint to our instinctive search for the disruptive effect of the exceptional artist or historical event.

Art historians have a much trickier relationship with spatial analysis, as we have a poor habit of confounding questions about place with questions about space. Cynthia actually presents a great example of clear thinking about this difference. As she demonstrates, the experience of the place Chichén Itzá is directly tied to movement and sight lines, which in turn depend on size, angle, and distance — that is, properties of space.

However, I charge everyone here to think carefully before you strike off to “make a map” just because you have latitude and longitude. Take this map that shows relative urban populations in Europe in 1650. We can create maps that happen to communicate relative populations… but just because these cities have coordinates does not mean that a map is necessarily the best way to think about their populations.

A chart can often be far more illuminating — here, we can compare the changing regional share of urban population between the sixteenth and eighteenth century. In this case, cities are not interesting as points on a map, but as places whose a-spatial identities are more important than their literal locations.

I also want to prod potential mappers to think about integrating networks into their research. As Saul Steinberg reminded us, the distance from one point to another can be as much about affective qualities as it is about the literal distance. These relationships can be thought of as a network that just happens to have spatial attributes.

The ORBIS project by Walter Scheidel and Elijah Meeks is a tour de force example of this kind of thinking. From an intricate model of the transport network available to travelers in the Roman world,they can re-project the shape of the empire as it may have felt from a myriad of perspectives: from Constantinople… versus Rome. This offers a new approach to thinking about centrality, periphery, and frontier zones.

Here, they compute a potential frontier between the eastern and western empire as the zone where travel time between the two capitals was equal. Construed as such, the frontier is no longer a static line, but instead would shift depending on the season, or on which means a traveler had at their disposal. What if Cynthia re-projected her terrestrial maps of Chichén Itzá to compute the affective distance between structures depending on the identity of the traveler, the time of day, or the weather? What if we applied this thinking to the installation of objects in an exhibition or museum, or the sequence of chapel altarpieces in a church? Consider carefully how you can merge topological, or network, questions, with topographical, or spatial ones.

The catch is that it is difficult, if not impossible, to do these more complex analyses with off-the-shelf software like Gephi, or even ArcGIS. And that is why I want to close by reflecting on how we could integrate these methodologies into training the next generation of scholars. Daniela has beautifully integrated a digital assignment into a course that was primarily focused on global architectural history. By using a novel method for composing and communicating their ideas, students were compelled to engage the subject material more attentively, and to be more self-aware about how they perform their scholarship. But what comes next for students who want to go further? Does Minnesota offer a course where students can read widely and deeply in literature on space and GIS in the context of architecture? Does the department view this approach as a fully-fledged methodology, rather than a simple exercise that can be added atop a core curriculum?

This leads me to my first main takeaway: pushing the envelope of our computational practice is as much about technical literacy as it is about learning a bibliography. This is a major investment, much like learning the literature for post-colonialist, feminist, or semiotic approaches to art history. This is why I want to resist thinking about network or spatial analysis as mere “tools”. These are entire disciplines, with histories, institutions, and theoretical disputes as rich as any in art history. Read sociologists and geographers grappling with these methods! Push your own art historical questions to take creative advantage of these methods. They offer far more than pretty graphs and maps.

And this isn’t just about language. We all work within organizations that love doing things by half measures and baby steps. “Let’s not do a full course on digital approaches, let’s just have a unit in a methods seminar.” Or, “Let’s just hire a postdoc, rather than a fully-fledged faculty line.” And I don’t say this to berate anyone within this room — I know the restraints that all of you work within. And believe me, I will gladly take that postdoc position! But we can build all the support programs, computer labs, and graduate assistantships we want: without a building up a corps of faculty members who engage deeply with these methods, and who have the purview to lead innovative research, we’ll never fulfill any of the promises that computing offers to art history. We need to start by giving it the full intellectual investment it deserves.