This weeks readings on network analysis represent the second week of readings on specific methodologies. Out of all five weeks (text analysis, network analysis, visualization, mapping, humanities computing) on methodologies, I have the least amount of experience with networks. Yet, in some ways, I find them the most fascinating. They are fascinating because the crux of networks are relationships. What are the relationships between certain “things?” Networks, in essence, are very similar to my field of study, mapping. While networks remove in large part the “geographic context,” both maps and networks focus on the relationship various “things” have between one another. While maps are inherently a spatial platform, networks can represent a plethora of relationships.
In all honesty, the reading that helped the most (as others in the course have mentioned as well) was Scott Weingart’s article in The Journal of Digital Humanities, “Demystifying Networks, Part 1 & 2.” While he writes for an audience with no experience or knowledge of networks, I found Weingart’s warnings to be particularly interesting. He warns of two pitfalls:
- When you learn a new technique, very quickly everything looks like it can be solved with this new technique. “When you learn to use a hammer, everything looks like a nail.”
- Methodology appropriation is dangerous as every methodology comes with a list of caveats to the user. Furthermore, when lifting that methodology out of its field and placing it into your own, those caveats are easily overlooks, forgotten, or just ignored.
I have noticed something of myself as we have started to work through the methodologies. I am particularly attuned or focused on finding the boundaries. Last week on text analysis, I was attuned to any mention of warnings or shortcomings of the methodologies int he reading. In addition, when we arrive at the mapping week (I am stoked for that week’s readings) I made sure to add Mark Monmonier’s How to Lie with Maps because I am well aware of the limits of that platform/methodology.
At any rate, Weingart’s second warning stuck with me throughout the rest of the readings. Methodologies appropriation is something that DH is particularly attune to while still overlooking it. This could lead into a discussion of the “black box” mentality of using tools to applying the wrong algorithm to the data set (Weingart’s explanation of applying a centrality analysis is particularly illustrative).
More importantly, and what I am still interested in delving into is the varying nature of data sets. Weingart notes that network analysis is not equipped to handle multi-modal analysis well. His article focuses around examples of single or bi-modal analysis. I am left wondering what is the recourse for historians whose data and sources are almost always heterogeneous? The article was published in 2011, 4 years ago, and perhaps network analysis has progressed to a point that multi-modal analyses can be dealt with much easier and with a greater degree of results.
Overall, what is the impact heterogeneous data has on using network analysis as a viable platform of research?