The use of computational methods to explore complex social and cultural phenomena is growing ever more common. Geographic information science in the service of better understanding the shape and scale of the Holocaust [1], natural language processing techniques leveraged to detect style and genre of 19th century literature, or the use of information visualization to present and interrogate each of these subjects is already happening. Among these techniques, it is the study of networks and how they grow that may be the most interesting.
Modern mathematical network analysis techniques have been around for decades, whether as developed to identify centrality in social networks or to distort topography to reflect topology in transportation networks, such as the work of geographer Waldo Tobler. But the growing accessibility of tools and software libraries to build, curate, and analyze networks, along with the growing prominence of such networks in our everyday lives, has lead to a wealth of applications in digital humanities and computational social sciences.
When we use networks to study culture and society, we perform an important shift in perspective away from the demographic and biographical to a focus on relationships. The study of networks is the study of the ties that bind people and places and objects, and the exhaustive details of those places and people, which are so important to traditional scholarship, are less important when they are viewed in a network. It is the strength and character of the bonds that define an actor’s place in a network, not the list of accomplishments that actor may have, though one would expect some correlation. In changing our perspective like this, we discover the nature of the larger system, and gain the ability to identify overlooked individuals and places that may have more prominence or power from a network perspective.
Many of the networks studied by researchers are social networks, with the historical kind being the most difficult to approximate and comprehend. Historical networks deal with difficult problems of modeling and representation. In the 16th century Spanish scientists shared geographic locations and subject matter of study, but some gamed the system and claimed connection to other, more prominent scientists or activity in fields that was not true. The China Biographical Database [2] has nearly 120,000 entries for Chinese civil servants, their kinship ties, their offices and posting, and the events in their lives, but only half have known affiliations. In the case of historical networks, the unevenness of the data may not be systematic, and it might even be the result of intentional misrepresentation.
Other networks are not social networks per se. In “ORBIS: The Stanford Geospatial Network Model of the Roman World” [3, 4], the goal was to build a parsimonious transportation network model of the Roman World with which to compile and better understand movement of people and goods in that period and region. To do so required not only the tracing of Roman roads using GIS, but the simulation of sailing to generate coastal and sea routes to fill out the network. The result of such a model is to provide the capacity to plan a trip from Constantinople to Londinium in March and see the cost according to Diocletian’s Edict and the time according to a schematic speed for the vehicle selected. But more than that, the ORBIS network model is an argument about the shape and nature of the Roman World, and embedded in it are claims such as that the distance of England from the rest of Rome was variable, and that changing the capital—moving the center of the network—would have systematic effects on the nature of political control.
Networks are inherently models that involve explicit, formal representation of the connection between individual elements in a system. But the accessibility of tools to represent and analyze such models has outstripped the familiarity with the methods for doing so. You can now calculate the Eigenvector Centrality of your network with the push of a button, but understanding what Eigenvector Centrality is still takes time and effort. More complex techniques for understanding the nature of networks, like the Exponential Random Graph Models being studied at the AIM workshop this week, require even more investment to understand and deploy. But the results of the use of computational methods in the exploration of history and culture are worth that investment.
It may be that information or data visualization will play a role in the greater adoption and understanding of these complex techniques. This is especially true as we move away from the static representation of data points and toward the visual representation of processes, such as Xueqiao Xu’s interactive visualization of network pathfinding [5]. Such visualizations make meaningful the processes and functions to audiences that may not be familiar with mathematical notation or programming languages. Scheidel, in his paper “The shape of the Roman World” utilizes dynamic distance cartograms—made possible as a result of creating a network—to express a Roman world view with a highly connected Mediterranean coastal core and inland frontiers. While this relatively straightforward transformation of geographic space to represent network distance could have been expressed with mathematical notation, data visualization is more accessible to a broader audience.
Networks are allied with notions of social power, diffusion, movement, and other behavior that have long been part of humanities and social science scholarship. The interconnected, emergent, and systematic nature of networks and network analysis is particularly exciting for the study of culture and society. Other computational methods do not so readily promote the creation of systems and models like networks do. But doing so will often require dealing with issues of uncertainty and missing evidence, especially in the case of historical networks, and require a better understanding of how networks grow and change over time. It will also require some degree of formal and explicit definition of connection that reflects fuzzy social and cultural concepts that, until now, have only been expressed in linear narrative.
References:
[1] The Spatial History Project: Holocaust Geographies, Stanford University
[2] China Biographical Database Project (CBDP), Harvard University
[3] ORBIS: The Stanford Geospatial Network Model of the Roman World
[4] Walter Scheidel, The shape of the Roman world (pdf)
[5] Xueqiao Xu, Pathfinding.js
Elijah Meeks
Digital Humanities Specialist
Stanford University