|
An experiment in writing in public, one page at a time, by S. Graham, I. Milligan, & S. Weingart

Networks Analysis

1 Leave a comment on paragraph 1 0 Networks and network visualizations are becoming increasingly important in digital humanities and digital history research. There are however very few resources meant for historians and the particular challenges that network analysis pose for us. In this chapter, we cover the fundamentals of network analysis so that the approach can be used in your own research – and you will know how to critically read others’ use of networks in theirs.

2 Leave a comment on paragraph 2 0 Formal networks are mathematical instantiations of the idea that entities and connections between them exist in consort. They embody the idea that connectivity is key in understanding how the world works, both at an individual and a global scale. Graph theory, social network analysis, network science, and related fields have a history dating back to the early eighteenth century, cropping up in bursts several times since then. We are currently enjoying one such resurgence, not incidentally co-developing along with the popularity of the Internet, a network backbone connecting much of the world to one system.

3 Leave a comment on paragraph 3 0 The idea that relationships are essential to understanding the world around us is, of course, ancient. The use of formal network methods for historical research, however, is much more recent, with only a few exceptions dating back beyond thirty years. Marten Düring has aggregated a thorough multilingual bibliography at http://historicalnetworkresearch.org for a list of specific instances. This chapter will go over a few examples of how historians have used networks, in what situations you might or might not want to use them, and the details of how networks work mathematically and technically.

4 Leave a comment on paragraph 4 0 In the 1960s, Eugene Garfield created the “historiograph”, a technique to visualize the history of scientific fields using a network of citations or historical narratives laid out temporally from top to bottom.[1] Garfield developed a method of creating historiographs algorithmically, and his contemporaries hoped the diagram would eventually be used frequently by historians. The idea was that historians could use these visuals to quickly get a grasp of the history of a discipline’s research trajectories, either for research purposes or as a quick summary in a publication.

5 Leave a comment on paragraph 5 0 A citation analysis by White and McCann looking at an eighteenth-century chemistry controversy took into account the hierarchical structure of scientific specialties.[2] The authors began with an assumption that if two authors both contributed to a field, the less prominent author would always get cited alongside the more prominent author, while the more prominent author would frequently be cited alone. One scientist is linked to another if they tend to be subordinate to (only cited alongside of) that other author. The resulting networks, called entailograms, proved particularly useful in showing the solidification of a chemical “paradigm” over a period of 35 years. Lavoisier begins as a lesser figure in 1760, and eventually becomes the most prominent chemist by 1795; by that time, most chemists who were cited at all were cited alongside Lavoisier. Following the entailogram over time reveals conflict and eventual resolution.

6 Leave a comment on paragraph 6 0 Citation analysts, also called bibliometricians or scientometricians, is still a rapidly growing discipline, but despite the hopes of its founders and though many if its practitioners conduct historical research, historians rarely engage with the field.

7 Leave a comment on paragraph 7 0 Historical sociologists, anthropologists, economists, and other social scientists have been using formal network methods for some time, and tend to exchange ideas with historians more readily. One such early sociological work, by Peter Harris, also employed citation analysis, but of a different sort. Harris analyzed citations among state supreme courts, looking at the interstate communication of precedent from 1870-1970.[3] By exploring who relied on whom for legal precedent over the century, Harris showed that authority was originally centralized in a few Eastern courts, but slowly became more diffuse across a wide swath of the United States. Network approaches can be particularly useful at disentangling the balance of power, either in a single period or over time. A network, however, is only as useful as its data are relevant or complete. We need to be extremely careful when analyzing networks not to read power relationships into data that may simply be imbalanced.

8 Leave a comment on paragraph 8 0 In a study on nineteenth century women’s reform in New York, Rosenthal et al. reveal three distinct periods of reform activity through an analysis of organizational affiliations of 202 women reform leaders.[4] These two hundred women were together members of over a thousand organizations, and the researchers linked two organizations together based on how many women belong to them both. The result was a network of organizations connected by the overlap in their member lists, and a clear view of the structure of women’s rights movements of the period, including which organizations were the most central. The study concludes, importantly, by comparing network-driven results to historians’ own hypotheses, comparing its strengths and weaknesses with theirs. For research on organizations, network analysis can provide insight on large-scale community structure that would normally take years of careful study to understand. As much as networks reveal communities, they also obscure more complex connections that exist outside of the immediate data being analyzed.

9 Leave a comment on paragraph 9 0 The study of correspondence and communication networks among historians dates back centuries, but its more formal analysis is much more recent. The Annales historian Robert Mandrou[5] and the historian of science Robert A. Hatch[6] both performed quantitative analyses of the Early Modern Republic of Letters, exploring the geographic and social diversity of scholars, but neither used formal network methods. In a formal network study of Cicero’s correspondence, Alexander and Danowski[7] make the point that large scale analyses allows the historian to question not whether something exists at all, but whether it exists frequently. In short, it allows the historian to abstract beyond individual instances to general trends. Their study looks in 280 letters written by Cicero; the network generated was not that of whom Cicero corresponded with, but of information generated from reading the letters themselves. Every time two people were mentioned as interacting with one another, a connection was made between them. Ultimately the authors derived 1,914 connections between 524 individuals. It was a representation of the social world as seen by Cicero. By categorizing all individuals into social roles, the authors were able to show that, contrary to earlier historians’ claims (but more in line with later historians), knights and senators occupied similar social/structural roles in Cicero’s time. This is an example of a paper that uses networks as quantitative support for a prevailing historical hypothesis regarding the structural position of a social group. Studies of this sort pave the way for more exploratory network analyses; if the analysis corroborates the consensus, then it is more likely to be trustworthy in situations where there is not yet a consensus.

10 Leave a comment on paragraph 10 0 In what is now a classic study (perhaps the only study in this set relatively well-known beyond its home discipline), Padgett and Ansell used networks deftly and subtly to build a historical hypothesis about how the Medici family rose to power in Florence.[8] The authors connected nearly a hundred 15th century Florentine elite families via nine types of relations, including family ties, economic partnerships, patronage relationships, and friendships. Their analyses reveals that, although the oligarch families were densely interconnected with one another, the Medici family – partially by design and partially through happy accident – managed to isolate the Florentine families from one another in order to act as the vital connective tissue between them. The Medici family harnessed the power of the economic, social, and political network to their advantage, creating structural holes and becoming the link between communities. Their place in the network made the family a swing vote in almost every situation, giving them a power that eventually gave rise to a three hundred year dynasty.

11 Leave a comment on paragraph 11 0 Before Facebook and MySpace, the first network of people to come to mind would probably kinship or genealogical networks with linkages between family members. Historiographic studies of these networks stem from early prosopographical (or collective biography) methods, but explorations using social network analysis are more recent. Looking at a large town in southwest Germany in the early nineteenth century, Lipp explored whether and how the addition of an electoral system affected the system of kinship networks that previously guided the structure of power in the community.[9] Surprisingly in an area to become known for its democratic reforms, Lipp showed that a half century of elections had not reduced the power of kinship in the community – in fact, kinship power only became stronger. Lipp also used the network to reveal the prominent actors of local political factions and how they connected individuals together. In this case, networks were the subject of study rather than used as evidence, in an effort to see the effects of political change on power structures.

12 Leave a comment on paragraph 12 0 Trade networks are particularly popular among economists, but have also had their share of historical studies. Using the records of nearly five thousand voyages taken by traders of the East India Company between 1601 and 1833, Erikson and Bearman show how a globalized economy formed out of ship captains seeking profit out of the malfeasance of private trade.[10] Captains profited by using company resources to perform off-schedule trades in the East, inadvertently changing the market from a dyadic East-West route to an integrated and complex global system of trading. The authors used a network as evidence, in this case the 26,000 links between ports each time two were connected along a trading route. Over two hundred years, as more ports became connected to one another, the East India Company lost control in a swath of local port-to-port connections. The authors show that the moments at which private trade were at its peak were also critical moments in the creation of more complex trade routes. While network analysis is particularly powerful in these many-century longitudinal studies, they also must be taken with a grain of salt. Without at least a second dataset of a different variety that is connected to the first, it is difficult to disentangle what effects were caused by the change in network structure, and what effects were merely external and changed both the network and the effect being measured. Global networks also tend to entangle geographic and relational distances, a fact which should not be glossed over when trying to understand the lived experiences of historical actors, which may diverge greatly from a network representation.

13 Leave a comment on paragraph 13 0 Folklorists have a long tradition of classifying folktales based on types, motifs, and various indices in order to make finding, relating, and situating those tales easier on the scholars balancing thousands upon thousands of tales. These schemes are often inadequate to represent the multidimensional nature of folktales, such as a tale that is classified as being about manor lords, but also happens to include ghosts and devils as well. Tangherlini and colleagues[11] came up with a solution by situating a collection of nineteenth century Danish folktales in a network that tied tales to subjects, authors, places, keywords, and the original classification schemes, resulting in a network connecting 3,000 entities together by 50,000 ties and made them easily browsable in an online interface. The interface made it significantly easier for folklorists to find the tales they were looking for. It also aided in serendipitous discovery, allowing scholars to browse many dimensions of relatedness when they were looking at particular tales or people or places.

14 Leave a comment on paragraph 14 0 Lineage studies with networks are not limited to those of kinship. Sigrist and Widmer[12] used a thousand eighteenth century botanists, tracing a network of between masters and disciples, to show how botany both grew autonomous from medical training and more territorial in character over a period of 130 years. The authors culled their group of botanists from various dictionaries and catalogues of scientific biography, and found by connecting masters to disciples, they saw botanists from different countries had very different training practices, and the number of botanists who traveled abroad to study decreased over time. The study juxtaposes a history of change in training practices and scientific communities against traditional large scientific narratives as a succession of discoveries and theories.

15 Leave a comment on paragraph 15 0 As is clear, historical network analysis can be used in a variety of situations and for a variety of reasons. The entities being connected can be articles, people, social groups, political parties, archaeological artefacts, stories, and cities; citations, friendships, people, affiliations, locations, keywords, and ship’s routes can connect them. The results of a network study can be used as an illustration, a research aid, evidence, a narrative, a classification scheme, and a tool for navigation or understanding.

16 Leave a comment on paragraph 16 0 The possibilities are many, but so too are the limitations. Networks are dangerous allies; their visualizations, called graphs, tend to be overused and little understood. Ben Fry, a leading voice in information visualization, aptly writes:

17 Leave a comment on paragraph 17 0 There is a tendency when using graphs to become smitten with one’s own data. Even though a graph of a few hundred nodes quickly becomes unreadable, it is often satisfying for the creator because the resulting figure is elegant and complex and may be subjectively beautiful, and the notion that the creator’s data is ‘complex’ fits just fine with the creator’s own interpretation of it. Graphs have a tendency of making a data set look sophisticated and important, without having solved the problem of enlightening the viewer.[13]

18 Leave a comment on paragraph 18 0 It is easy to become hypnotized by the complexity of a network, to succumb to the desire of connecting everything and, in so doing, learning nothing. The following chapter, beyond teaching the basics of what networks are and how to use them, will also cover some of the many situations where networks are completely inappropriate solutions to a problem. In the end, the best defense against over- or improperly- using a network is knowledge; if you know how the ins and outs of networks, you can judge how best to use them in your research.



19 Leave a comment on paragraph 19 0 [1] Eugene Garfield, “Historiographs, Librarianship, and the History of Science.” In Toward a Theory of Librarianship: Papers in Honor of Jesse Hauk Shera, edited by Conrad H. Rawski, 380–402. (Metuchon, NJ: Scarecrow Press, 1973).

20 Leave a comment on paragraph 20 0  

21 Leave a comment on paragraph 21 0 [2] Douglas R. White and H. Gilman McCann, “Cites and Fights: Material Entailment Analysis of the Eighteenth-Century Chemical Revolution.” In Social Structures: A Network Approach, by Barry Wellman and Stephen D. Berkowitz, 380–400 (Cambridge, UK: Cambridge University Press, 1988).

22 Leave a comment on paragraph 22 0  

23 Leave a comment on paragraph 23 0 [3] Peter Harris, “Structural Change in the Communication of Precedent among State Supreme Courts, 1870–1970.” Social Networks 4.3 (1982): 201–12..

24 Leave a comment on paragraph 24 0 [4] Naomi Rosenthal, Meryl Fingrutd, Michele Ethier, Roberta Karant, and David McDonald.“Social Movements and Network Analysis: A Case Study of Nineteenth-Century Women’s Reform in New York State.” American Journal of Sociology 90.5 (1985): 1022–54.

25 Leave a comment on paragraph 25 0 [5] Robert Mandrou. From Humanism to Science 1480-1700, 2nd Ed. (Atlantic Highlands, NJ: Humanities Press, 1978).

26 Leave a comment on paragraph 26 0 [6] Robert Alan Hatch, “Between Erudition & Science: The Archive & Correspondence Network of Ismaël Boulliau.” In Archives of the Scientific Revolution: The Formation and Exchange of Ideas in Seventeenth-Century Europe, edited by Michael Cyril William Hunter (Woodbridge: Boydell & Brewer, 1998).

27 Leave a comment on paragraph 27 0 [7],Michael C. Alexander and James A. Danowski,.“Analysis of an Ancient Network: Personal Communication and the Study of Social Structure in a Past Society.” Social Networks 12.4 (1990): 313–35.

28 Leave a comment on paragraph 28 0 [8] John F. Padgett and Christopher K. Ansell, “Robust Action and the Rise of the Medici, 1400-1434.” American Journal of Sociology 98.6 (1993): 1259–1319.

29 Leave a comment on paragraph 29 0 [9] Carola Lipp, “Kinship Networks, Local Government, and Elections in a Town in Southwest Germany, 1800-1850.” Journal of Family History 30.4 (2005): 347–65..

30 Leave a comment on paragraph 30 0 [10] Emily Erikson and Peter Bearman, “Malfeasance and the Foundations for Global Trade: The Structure of English Trade in the East Indies, 1601–1833.” American Journal of Sociology 112.1 (2006): 195–230.

31 Leave a comment on paragraph 31 0 [11] James Abello, Peter Broadwell, and Timothy R. Tangherlini,. “Computational Folkloristics.” Communications of the ACM 55.7 (2012): 60..

32 Leave a comment on paragraph 32 0 [12] René Sigrist and Eric D. Widmer, “Training Links and Transmission of Knowledge in 18th Century Botany : A Social Network Analysis.” Redes: Revista Hispana Para El Análisis de Redes Sociales 21.0 (2012): 347–87.

33 Leave a comment on paragraph 33 0 [13] Ben Fry, Visualizing Data: Exploring and Explaining Data with the Processing Environment (Sebastopol, CA: O’Reilly Media, 2007).

Page 57

Source: http://www.themacroscope.org/?page_id=889