Creating data from humanities sources – if done with care – can enable scholars to identify and potentially answer a number of questions that purely qualitative analyses alone cannot (or at least, for which qualitative analyses are less well-suited). With an extensive enough data set, humanities researchers can engage in statistical, networking, and mapping analyses that can identify and evidence broad patterns not readily apparent in the smaller data collections with which qualitative studies are typically engaged. These patterns will still need to be placed in the context of the scholar’s nuanced knowledge of their subject area in order to be meaningful, but the impetus and clarity provided by digital data can be very helpful. (Posner, 2015)
For instance, the “Tudor Networks of Power” project was able to map networks of official influence across Britain and Western Europe during the Tudor era using data relating to the velocity (i.e. frequency) of royal and diplomatic correspondence. The data was derived from the Tudor State Papers collection, which had been digitally archived via the State Papers Online but not transformed into data models usable for large-scale analysis. By closely examining and disambiguating XML data related to the collection, the project scholars were able to identify at least 20,566 unique individuals corresponding across 5,409 unique geographic locations; this data was then used to map the frequency of correspondence between these nodes. By viewing these networks, researchers were then able to pose new questions, or even tentatively suggest hypotheses regarding the exercise of power and influence across the Tudor kingdom. (Ahnert and Ahnert, 2023)
Similarly, in our own class project tracking various data points across the narratives of former American slaves (or, more ambitiously, in the similar project of Paul Escott examining a much larger data set), we were able to begin searching for patterns and correlations among the life experiences of various former slaves across the American South. While smaller-scale comparative studies such are common to comparative historical research, having a large data set of multiple former slave narratives, appropriately coded for particular data models, can allow researchers to identify larger-scale patterns not readily apparent (or at least statistically demonstrable) from smaller-scale studies.
Of course, as with any argument, the conclusions will only be reliable if they are derived from valid premises. In digital humanities data analysis, care must be taken along each step in gathering sources, compiling them into data models, and drawing appropriate conclusions. The process of transforming sources into data is not merely mechanical but a fundamentally interpretive one. Even before the scholar has begun gathering their sources, crucial decisions may already have been made for them decades or even centuries earlier – choices such as what records to retain for posterity, what types of information to include in those records, how to organize those records, and even where to house them (collections can be broken up by intent or happenstance over the years). Gaps and biases in the very creation and assembling of the original sources must always be taken into account. (Drucker 2021) If working with digital sources, the scholar must also be mindful that only a small selection of records tend to be digitized, and the decision of which records to prioritize for the digitization process is inherently determined by the needs of administrators and donors, i.e. it is a political decision. And at the point of converting sources in data points, scholars are confronted with innumerable interpretive and even ethical decisions. Given the vast amount of possible data points in, say, just a single former slave narrative, which points will the scholar choose to code into their spreadsheet (or other coding medium of choice)? How will the messiness and ambiguity of reality be reduced into a relatively small range of fixed values conducive to modeling, or even a binary TRUE/FALSE field? Should the “raw” data found in the sources be taken at face value, or should the scholar exercise discretion to account for possible biases or inaccuracies at the point of the record’s creation?
These considerations by no means render the practice of digital humanities data analysis too unreliable for use, but they should warn us against undue faith in the “neutrality” of seemingly mechanical or purely mathematical models. The best practice for humanities scholars engaged in digital data analysis is to make transparent the various points along the chain where distortions may have crept in. What process was used to create the “raw” records in the first place – whose viewpoint do they reflect, what power relationships were involved in their creation, and were they subject to alteration before being archived in their “final” form? What records were not created and therefore did not find their way into this collection? How complete was the original archive, and what gaps in the records exist? (Hoekstra and Koolen, 2019). What choices did the project scholars make when selecting their raw data, and what was the reasoning for the data models they created from it? What was their approach to disambiguating the inevitable points of confusion in any large data set? How should the arguments we derive from this data be limited or qualified by our answers to these other questions? Any responsible use of data analysis in the digital humanities will bear all of these considerations in mind.
Bibliography
Ahnert, Ruth and Sebastian Ahnert “Tudor Letters in the Digital Age.” In Tudor Networks of Power. Oxford: Oxford University Press, 2023. https://academic-oup-com.mutex.gmu.edu/book/51646/chapter/419632240.
Drucker, Johanna. “Data Modeling and Use.” In The Digital Humanities Coursebook: An Introduction to Digital Methods for Research and Scholarship, 1st ed. Taylor and Francis, 2021.
Hoekstra, Rick, and Marijn Koolen. “Data scopes for digital history research.” Historical Methods: A Journal of Quantitative and Interdisciplinary History 52, 2 (2019): 79-94.
Posner, Miriam. “Humanities Data: A Necessary Contradiction.” Miriam Posner’s Blog, June 25, 2015. https://miriamposner.com/blog/humanities-data-a-necessary-contradiction/.