This page is referenced by:

• 1 2019-09-25T09:07:26-07:00 The Validity of Network Analysis 10 An examination of the effectiveness of network analysis within CLS. plain 2019-09-27T21:55:20-07:00 by Ezequiel Joaquin Torres
  
      Digital Humanities (DH) is a field that concerns itself with the presence of human generated information and cultural works within the emerging digital realm. In recent decades, centuries of work once exclusively stored and communicated through books, papers, journals, and manuscripts are now finding a new technological channel in digital archives and electronic communications. This has prompted the evolution of DH. Computational Literary Studies (CLS) is a subfield within DH that aims to analyze these literary records and communications with digital tools. Through analytical scientific methods, CLS intends to provide empirical data that may form the basis for insights previously unnoticed through scholarly discourse.
      Network Analysis (NA) is a tool that is often used within CLS. It represents a set of objects and their relevant information as a set of nodes which are connected and mapped/positioned according to a set of rules. Books can be represented as nodes and mapped via genre, as could paragraphs within a text via their tone. The resultant network map shows the web of connections between nodes, and can reveal patterns present throughout the entire set of objects that might not have been noticed without network visualization.
      Network analysis, along with the field of CLS as a whole, is oftentimes criticized as ineffective and impractical. Some claim that NA lacks scientific rigor, and that the information that it gleans is statistically useless. This is true, to an extent; judging by various examples of studies conducted within the digital humanities that use network analysis, the results can frequently use improvement. However, these examples are not a practical lense which to judge NA in its entirety. Network analysis itself is not an inherently useless tool, as many critics seem to believe; rather, its utility must be preserved through careful application within the context of CLS.
      Nan Z. Da, an English professor at the University of Notre Dame, is one such critic. In her perspective, CLS receives too much funding for too little practical results. She claims that, for a purportedly scientific field, CLS has little scientific rigor or statistical math applied to its results (Da 610). In her paper, she categorizes a set of CLS studies into two classes. One class is composed of studies whose results prove statistically strong and reliable, but make points that are obvious and unhelpful in reaching a deeper analysis of text unattainable by non-computational means. The other class is filled with studies which make innovative and productive results, but are statistically weak and untrustworthy (Da 607).
      Among the studies that Da categorizes as obvious and unhelpful is one by Ed Finn that claims to perform a network analysis centered around Junot Diaz’s The Wondrous Life of Oscar Wao’s Amazon page. Finn’s study mapped out book recommendations on each book’s page as a node in a network, to seemingly prove that Junot Diaz’s book is a gateway that links two different academic discourses (Finn 2013). Da claims that this study is a network analysis that lacks any sort of true analysis, or even accurate portrayal as a network (Da 631). A major flaw she points out is that the edges connecting the nodes don’t give indication to the weight of their connection. According to her, the true complexity of the exchange of ideas through literature cannot be represented in such a manner (Da 632).
      Though not directly concerned with CLS, Mushon Zer-Aviv’s analysis of the use of NA echoes Da’s criticisms to an extent. He spoke about the unseen intricacies of the networks that increasingly dominate our manner of understanding of the world around us, from road maps and detective work to social media maps and representations of neural links. He warned of how our superficial grasp of network diagrams can lead to flaws in our comprehension of situations (Zer-Aviv, 2016). The biggest problem mentioned with networks was the lack of understanding of flow - that, by representing a system as a finite set of points and edges, it gives the impression that there’s nothing more to the system than that. There is no information conveyed about which way information flows from one node to another, by what means and protocols, and how it all interacts as a whole (Zer-Aviv, 2016). Zer-Aviv noted that what oftentimes is absent in networks can be the most useful: the protocols that dictate the manner of connection.

  Not emphasising the visualisation of the flow implies that only the layout of nodes and edges is enough to tell the whole story.
  ...
  Networks need narrative, both as a layer of annotation and as a way to present exemplary network flow. (Zer-Aviv, 2016)

      On top of that, there’s the growing tendency for networks to grow larger and expansive, to the extent that computers have to handle interpreting it all because no mere human could ever hope to comprehend it (Zer-Aviv, 2016). He claims that this can bring an illusion of a better understanding of a network, when in fact, it could just lead to pointless dead ends.
      In the end, Zer-Aviv’s message is nothing more than an effective warning about how we should carefully approach how we apply network analysis to everything. The title of his essay provides a succinct culmination of his thoughts; that is, if we are too careless with how we apply networks, they may cease to have any functional value overall. “If everything is a network, then nothing is a network” (Zer-Aviv, 2016).
      Despite the excessively negative tone and bias that is present throughout Da’s writing, her case against Finn’s study aligns with Zer-Aviv’s general perspective on network analysis. The problem with network analysis performed within CLS isn’t that the technique itself is flawed - rather, the problem is that no one seems to apply network analysis correctly.
      However, Da pessimistically believes that there is no correct manner of application for network analysis, since CLS can’t ever hope to quantify and measure the humanistic aspects of literature (Da 632). In contrast, Zer-Aviv doesn’t give up on the use of network analysis outright, though his perspective is agnostic to the field of CLS as a whole.
      As Zer-Aviv points out, perhaps the flaw is that network analysis is often used alone (Zer-Aviv, 2016). It isn’t a weak technique because it can’t represent all aspects of a situation - nay, if anything, its potential is weakened by all attempts to make network analysis a tool that can surpass the rich narrative behind the situations it represents (Zer-Aviv, 2016). Subsequently, the real issue with its application is that researchers continue to try to draw data from static networks, such as Finn’s elementary diagram of connected books, which are limited in capability and scope due to the simplicity forced upon them by their rigorously statistic-compliant design. Networks that embrace and acknowledge the dynamic nature of the humanities are much more useful than many people seem to believe. Researchers, in their excitement to provide precise analytical information, forgo creativity that could prove insightful.
      Critics may oftentimes attack this as a point of weakness. Nan Da does this with the group of studies she labelled intriguing, but statistically useless (Da 607). However, when one considers that the point of the analysis isn’t to provide hard facts, this claim becomes frivolous rather quickly. Their measures and statistics can be useful even if they can’t be proven entirely correct. It is being applied within the humanities, after all - a field that thrives upon analysis in the realm of material highly influenced by the fluid nature of human perspectives and ideas. Instead of shunning network analysis for failing to be a scientifically rigorous tool, the humanities should accept network analysis for what it is - a tool that can be used to provide a window into novel perspectives previously unseen.
  
  
  Works Cited
      Da, Nan Z. “The Computational Case against Computational Literary Studies” Critical Inquiry, vol. 45, no. 3, Mar 2019, pp.601-639. https://doi.org/10.1086/702594
      Finn, Ed. “Revenge of the Nerd: Junot Diaz and the Networks of American Literary Imagination,” Digital Humanities Quarterly 7, no. 1, 2013. www.digitalhumanities.org/dhq/vol/7/1/000148/000148.html
      Zer-Aviv, Mushon. “If everything is a network, nothing is a network” Visualising Information for Advocacy, Jan 8 2016. https://visualisingadvocacy.org/node/739.html