Margaret K. Smith
March 16, 2021
Null Results in the Digital Humanities
Sometimes digital humanities projects don’t work. That’s not an indictment of the scholar or the research question. It’s an unavoidable facet of taking humanistic data that is often fragmentary, contradictory, misleading, or simply opaque, and attempting to order it in a way that computers can usefully parse. It’s a process that is by nature experimental and iterative.
But the nature of academic scholarship is that those failed projects don’t see the light of day for any number of reasons. Perhaps it’s a desire to hide our dirty laundry, but more often it’s simply a matter of time and energy – there’s very little incentive to give more hours to a project that’s already yielded very little for the untold hours we’ve spent on it. Onward and upward, et cetera.
When we only see the projects that work – the ones that produce attractive visualizations or compelling stories, the ones that overturn established narratives or uncover silenced voices, the projects that change the field or capture a classroom – we receive a fragmentary and misleading view of digital humanities scholarship. Humanistic data isn’t neat, and the ways that we try to tidy it up matter both for our individual projects and for the field at large. Every step you take to encode your data is an argument, a component part of your broader thesis. And a failed project doesn’t mean that your broader thesis is bad – it just means that somewhere along the way, one of those arguments was based on an assumption that didn’t hold true. Sometimes the successful projects have made the same assumption, and the project works anyway, so the problem goes undetected. A broken project doesn’t mean a bad project, and working visualizations don’t always mean good data.
I’ve often thought that digital humanities needs a sort of journal of null results, of the sort that scientific fields have begun to dabble with. We need a space to hash out the methodological challenges, the missteps, and the promising ideas that have yet to yield any promising outputs. In the absence of such a space, I thought I’d offer up one of my early failed projects, along with some lessons learned and thoughts for future endeavors.
Lines in the Landscape: The Expansion and Contraction of the Mac Carthaigh Riabhach
This project began its life as a paper proposal for the 2017 International Congress on Medieval Studies. I’d recently started working in earnest on my dissertation, which explored the strategies of negotiation employed by the MacCarthy Reagh lords of Carbery, Co. Cork, from 1366-1594. One of my chapters was going to focus on uses of the landscape, and as I gathered sources and data for that chapter, I grew frustrated with maps of the lordship that presented a fixed moment in time that masked all the fluid changes happening in what was a very contentious period. Or worse, they drew on lots of different sources and sort of silently elided them, presenting a composite image that never actually existed. So here’s my abstract, in which I proposed a dynamic visualization of the fluid expansion and contraction of the lordship:
“Following the formation of the Mac Carthaigh Riabhach as a distinct sept in the mid-thirteenth century, the new dynasty rapidly took its place among the most prominent lordships of later medieval Cork. From their territorial base in Carbery, the Mac Carthaigh Riabhach extended their influence through intermarriage, military alliance, and religious patronage. The expansion of influence was matched by territorial expansion throughout much of the fourteenth and fifteenth centuries. The tumult of the sixteenth and seventeenth centuries, however, drew the Mac Carthaigh Riabhach into several conflicts (sometimes as royalists, sometimes as rebels) and culminated in the loss of much of their territory after the 1641 Rebellion.
This paper uses digital tools to visualize the expansion and contraction of the lordship of the Mac Carthaigh Riabhach through its territorial extent and its social networks. Building on the genealogical research of John V. Kelleher, supplemented with historic maps, annals references, archaeological evidence and other primary sources, this paper offers a dynamic representation of the lordship of the Mac Carthaigh Riabhach as they responded to the changing environment of late medieval and early modern Cork. This project not only sheds light on the physical extent of the lordship, but also gives greater clarity and nuance to our understanding of the political position of the Mac Carthaith Riabhach throughout this period.”
So that’s ambitious. As it turns out, this is a much better synopsis of my entire dissertation than it is even of the landscape chapter, let alone a single conference paper. But it was my first foray into the dissertation after getting my prospectus approved, and I didn’t exactly know what I was dealing with yet.
I began by creating some base layers for specific moments: the pre-Norman kingdoms of the region as delineated in an early genealogical tract, the cantreds (administrative divisions) that constituted Anglo-Norman Ireland, and some Anglo-Norman-held lands as described in a 14th c. document, along with the boundaries of the townlands, parishes, and baronies that make up modern Cork.
Even this initial phase had its problems. There are no freely available shapefiles for medieval Irish boundaries, so I relied heavily on modern ones. These modern boundaries have changed relatively little since the Middle Ages, but they cannot be adopted wholesale without some very significant caveats. And in cases where it seemed like information was missing, I had filled in the gaps in what might generously be called an ad hoc fashion. Of course I had reasons for all of my interventions (although I doubt I could recreate them all now), but there was no easy way to identify those interventions on the maps I produced.
Nevertheless, with these layers established, I began creating a dataset of events that documented changes in the boundaries or extent of the MacCarthy Reagh lordship. There were methodological challenges here, too, primarily with regards to cleaning the data. For each item, I needed at least four pieces of data: a description of the event itself, a location, a date, and the actors. But reducing a vague or fragmentary historical text to clean and precise data often gave rise to issues of false precision. For instance, when Domhnall MacCarthy demolishes castles in Imokilly, but the annals don’t tell us which ones, how does that get reduced to coordinates for displaying on a map? Should I show all the castles I know to have existed in Imokilly in 1196? Should I mark the conflict at the border to demonstrate the MacCarthys active beyond their own territory? Should I pin it to the center of Imokilly? In this instance, I chose the last option, but I could have justified any of those choices. The same problem applies to dates. I can’t say with any certainty when Kilfeakle Castle was built. But I know that it exists when Domhnall attacks it in 1196, so in order to have it on my map’s time visualization, I have to give it a construction date of 1190, when the timeline begins. I know that I’ve made that intervention and why – but again, there’s no easy way to make that transparent on the resulting visualization.
Other problems were less about cleaning the data and more about the nature of my argument. I attempted to frame the narrative of my map in terms of ethnic identity. That’s a tricky maneuver even in a written argument, where you can footnote and caveat to your heart’s content. But on a map that might be stripped of its context and asked to stand alone? It’s a methodological hornet’s nest, and another one of those questions that turned out to be an entire dissertation (and more). In the end, I settled for a binary of “Irish” and “foreigner.” That might have flown for the early periods, since I started the visualization less than twenty years after the initial conquest. But since my project aimed to cover several centuries, a period marked by myriad forms of cultural hybridity and for which scholars continue to argue vociferously about ethnic identities, that binary broke down rapidly. So I tried encoding families and lordships. That was better from a data standpoint, but still plagued by ambiguous data points and much harder to visualize with the tools I had at hand.
I did end up producing a visualization, but it was neither the output I’d promised in my ambitious abstract, nor a piece of scholarship I was willing to stand by. So instead, my paper became a reflection on the challenges of this sort of research in the context of medieval Ireland and some thoughts for improving future projects.
My biggest takeaway from this project is that data transparency is absolutely imperative for digital humanities scholarship. This project highlights three facets of data transparency: how we deal with ambiguous data, where we’re eliding things for the purposes of analysis or visualization, and the basic inferences and assumptions on which we build our data sets. Every step we take when working with humanistic data is an argument, and just like we’d footnote or expound in an article or a book, we need to provide ways for colleagues to see our thought process, challenge our conclusions, or build on our work.
Some data sources (but not all, because I never wrote this up – a cautionary tale for another day!)
John O’Donovan, ed. and trans., Annals of the Four Masters (Dublin: Hodges and Smith, 1848-1851), volumes 3 and 4, electronic editions compiled by Myriam Priour, Floortje Hondelink, and Orla McDonald
Seán Mac Airt, ed., The Annals of Inisfallen, MS Rawlinson B 503 (Dublin: Dublin Institute for Advanced Studies, 1951), electronic edition compiled by Beatrix Färber, https://celt.ucc.ie//published/T100004/index.html
Paul MacCotter, Medieval Ireland (Dublin: Four Courts, 2008)
Liam Ó Buachalla, “An Early Fourteenth-Century Placename List for Anglo-Norman Cork,” Dinnseanchas 2, no. 1 (1966): 1-12, 39-66
Donnchadh Ó Corráin, “Corcu Loígde Land and Families” in Patrick O’Flanagan and Cornelius G. Buttimer, eds., Cork History and Society (Dublin: Geography Publications, 1993): 63-82