Anne Donlon

Black diasporic cultural histories can be imagined as a series of networks—of friendship, correspondence, political organizing, collaboration, travel, performance, publication, and education. To investigate these histories, scholars of the Black diaspora look to materials that have often been devalued and discarded or else inaccessible to the communities who produced them. People researching minoritized subjects not overtly represented in collections have learned to read between the lines and against the grain to find their subjects. Of course, there have also been collectors, librarians, and archivists who set out to create collections that correct for these omissions. They sometimes revised and refused the mainstream ways of describing collections, creating their own local vocabularies and systems for access when traditional professional standards were not sufficient.1 Typically, the organizational logic that structures archives is hierarchical; mainstream archival practices developed from the preservation of the records of powerful institutions and important white men. Archivists have had to imagine other options to counter these ideological structures—for instance, Wendy Duff and Verne Harris posit a “liberatory descriptive standard.”2

Digital methods offer possibilities for representing relations in new ways. Network graphs suggest a model to imagine relationships between historical and cultural materials in a less hierarchical way that would allow us to analyze and document a Black Atlantic past. But the data we use to create those networks and the points we are able to collect depend on the tools and documents to which we have access.

In this chapter, I discuss an effort to explore African American cultural networks by using archival finding aids to create network graphs. Digital methods and tools might offer increased access and possibilities for understanding archival collections in new ways, but they do not do so inherently. As the project failed to develop the way I had imagined, I began to turn my attention to the particular archival processes involved in creating the finding aids, as well as to the relevant institutional histories. I began to consider how archives could document them and how researchers could better understand how the interventions of archivists affect their experiences in archives. Finding aids can provide data to investigate Black diasporic cultural histories, but their structure and composition create challenges for visualizing relationships across collections. As libraries and archives make data available, reusable, and interoperable, there remain fundamental and epistemological questions about how such data both structures and obscures the stakes of what we are able to know.

Finding aids, like many things, reflect the biases and power structures that surround them. In the early 2000s, a body of literature emerged on finding aids, noting their social construction and imagining forms beyond the traditional finding aid.3 Today, conversations about library collections as data have made questions about the structure of finding aids again relevant.4 In addition, conversations about the racial demographics (in particular, the overwhelming whiteness) of archivists often highlight the impact of underrepresentation on the constitution and description of collections. Michelle Caswell calls to “dismantle white supremacy in archives,” noting “the paradox of a field that claims to value diversity but persists in replicating whiteness in demographics, values, and practices.”5 For archives that document the Black Atlantic and other archives of the African diaspora, we have to question how racism and colonialism affect archival description. Marisa Parham has asked, “We are left to wonder what new ways we might constitute the archive of black experience in the New World. From where is such data collected? How is it processed? How [do] we articulate an experience of data?”6 In this chapter, I apply these questions to metadata. What could we learn from metadata describing archival collections related to the Black experience? How can researchers better understand the tools we use to discover archival materials and how they structure our encounters with these historical materials?

During my time as a postdoctoral fellow at Emory University Libraries, I undertook a project with colleagues to create network graphs based on fifty-six African American collections at the Stuart A. Rose Manuscript, Archives, and Rare Book Library.7 I hoped that this method would allow us to reveal those hidden or forgotten figures that play central roles in cultural networks but are not themselves the central figure whose papers are collected. From my earlier dissertation research at the Rose Library, I knew of the letters and material in several collections that documented the friendship of Langston Hughes, Louise Thompson Patterson, Matt Crawford, and Evelyn Crawford and referenced their travels to the Soviet Union, as well as countries in Europe, the Caribbean, and Africa. As I familiarized myself with more collections, I found evidence of friendship, collaborations, and political organizing. I wanted to know more about the figures who themselves were not particularly famous but who played crucial roles in the production of twentieth-century Black Atlantic culture. I imagined that a network graph based on unpublished archival materials could illuminate a rich web of cultural and political relationships.

A number of archivists, developers, and digital humanists in the library at Emory were already engaged in efforts to make archival materials available and usable for digital scholarship. They were considering ways in which researchers might want to engage with digitized material, born-digital collections, or metadata made available in electronic form.8 For archival collections that are not born digital and have not been digitized, an electronic finding aid—a description of the contents of a collection created to help researchers access it—can offer a source of data to represent what is in the collection. This was an insight explored in the Belfast Group Poetry | Networks project, and which we adapted in the project with African American collections. However, as I soon learned, the format of a finding aid presents some challenges.

The level of description in a finding aid can vary widely across archives and even within an institution. Some finding aids describe every single item, others describe each folder, and still others describe series or groups of folders or boxes. I came to understand that this set of African American collections are described unevenly (or some might say flexibly). The structure and level of detail in a finding aid depend on the size and shape of the collection, the person processing it, the amount of time and resources allotted to the collection, and the era when it was processed (because practices have shifted over time). Although finding aids have some potential to serve as digital proxies for the collections themselves, archival description practices (and the variation therein) proved a challenge. Moreover, the hierarchical structure of finding aids suppresses some of the complexity represented in collections.

Although library collections as data hold a great deal of promise and appeal, the colonialist, racist legacies of many library practices must be acknowledged and challenged as we develop digital tools and processes for analysis. The editors of a 2015 issue of Social Text describe “the generative tension between recovery as an imperative . . . and the impossibility of recovery when engaged with archives whose very assembly and organization occlude certain historical subjects.”9 Black Studies scholars both recover devalued sources and draw attention to the ways historical narratives are constructed through the selection and erasure of evidence. To develop digital projects around the Black Atlantic using library materials, we must hold onto insights from those working in fields including Black digital humanities, Black Studies, and critical librarianship (#critlib). These scholars question what counts as knowledge and who controls the systems that codify knowledge. As Kim Gallon writes, “The black digital humanities help to unmask the racialized systems of power at work in how we understand the digital humanities as a field and utilize its associated techniques.”10 Together with critical library studies, Black digital humanities can take (and is taking) part in a conversation about how to dismantle inequities in libraries and archives.

Library metadata offers a proxy for materials that are not themselves in a digital form and that may never be in a digital form because of cost and other obstacles to digitizing. When I arrived at Emory, web developer Rebecca Sutton Koeser, digital humanities strategist Brian Croxall, processing archivist and metadata specialist Elizabeth Russey Roke, and others were completing a project that used collections’ finding aids (published as EAD, or Encoded Archival Description, an XML language) to create network graphs from a set of Rose Library collections related to Northern Irish poetry. In Belfast Group Poetry | Networks, the analysis focused on the biographical note and correspondence series in the finding aids of about ten Northern Irish poetry collections. In the project I discuss here, Koeser, Roke, and I extended that methodology to include other parts of the finding aid and applied it to a selection of fifty-six collections related to African American culture.

African American history and culture is one of the major collecting areas of Emory’s Rose Library and has been since the 1990s, when faculty and students pushed the university to commit resources to support the study of Black history and culture. Emory is a well-resourced, predominantly white university located in a predominantly white part of Atlanta. The university hired Randall Burkett as curator for African American collections in 1997, and the Rose Library has since become a major collector of African American materials.11 In this chapter I use the descriptor “African American” to describe these collections, in keeping with the Rose Library’s description of the collecting focus, whereas in more general discussion, I refer more broadly to Black diasporic collections and history.

In applying the methodology developed for the Belfast Group Poetry | Networks project to a larger set of African American collections, I came up against a shift in description practice that happened over time at Emory, which affected the shape of the networks and, in turn, prompted me to examine the racial politics of metadata and linked open data more closely. Most of the collections associated with the Belfast Group were collected and described at an earlier period, when the common practice at Emory was to create quite detailed finding aids. The Northern Irish poets’ finding aids were rich with names of their correspondents, which made for an informative network graph. Not every Irish poet’s collection was described in this way, which Croxall and Koeser discuss in “Archival Biases and Futures,” but the discrepancy became more pronounced when we examined the set of African American collections.12 The level of description in African American collections at the Rose Library vary: the earlier ones are described in a similar level of detail to the Northern Irish poetry collections, but over time the practice shifted, as I discuss in more detail later.

The purpose of a finding aid is not to provide an exhaustive index of the collection’s contents. Rather, it is supposed to help researchers locate collections of interest when they visit the reading room. In an interview, Roke noted that the finding aid’s purpose is to help a researcher “understand this person and the context around them” and not “to tell you exactly where to find that piece of paper containing that topic that you’re interested in. That’s kind of up to the researcher now.”13 Dorothy Berry, writing about the challenges of implementing Umbra Search African American History (umbrasearch.org), notes the difference between what kind of description is needed for machine use compared to what researchers need. Berry writes, “Description sufficient to promote use in a physical setting can be drastically different from description sufficient to promote use by both computer systems looking for records to aggregate and use by researchers searching for individual records in mass digital settings.”14 For Berry’s task, locating African American materials within larger collections to include in Umbra, the existing metadata was insufficient. In the network graph project, I encountered a similar obstacle to my original vision of discovering the important (and maybe hidden) figures who appeared across African American collections. The level of description in finding aids intended for in-person, physical use—particularly after a shift in processing practices at the Rose Library—was insufficient for the kind of analysis I initially envisioned.

As collaborations between digital humanists, archivists, and librarians continue to develop, the field should revisit the discussion of finding aids to consider what formats, addenda, or alternatives could make for more transparent, usable information. In the early 2000s, a number of archival scholars undertook critical analysis of the finding aid, grappling with the potential of digital formats and proposing ways to address bias. For instance, in “Colophons and Annotations: New Directions for the Finding Aid,” Michelle Light and Tom Hyry advocate for the inclusion of a colophon, an addendum with information about the processor’s interventions in arranging and describing the collection. The colophon would provide space for archivists to “acknowledge and explain their impact.”15

Duff and Harris grapple with issues of bias and complicity, proposing that “a liberatory descriptive standard would not seek to hide the movements of its construction.”16 Such a standard would “as far as possible, make known the biases of its creators” and encourage “the documentation of continuing archival intervention.”17 This attention to the role of the processing archivist and the ways that the archivist’s experience affects the description of collections can be further informed by contemporary conversations about the underrepresentation of archivists of color and of community archives.

Elizabeth Yakel’s “Archival Representation” also considers how finding aids and other tools can account for archivists’ interventions. In addition, Yakel questions whether electronic formats could be used more effectively and laments that EAD uses a traditional analog structure, “inhibit[ing] creative use of networked information and the emergence of new digital representational forms for the representation of primary sources.”18 Yakel’s critique of EAD and curiosity about the possibility of other digital representations are particularly relevant to evaluating this project. I was pushed to critically examine the structure and descriptive practices in finding aids when my project of finding aids networks fumbled. Whereas the Belfast Poetry Group | Networks project worked with a small group of interconnected and more evenly described collections, my larger undertaking did not exactly scale.

As mentioned previously, the density of naming within Rose Library finding aids has changed over time, in part reflecting shifts in the profession’s approach to description. In 2005, an article by Mark Greene and Dennis Meissner prompted a vigorous conversation about archival processing. They advocated “More Product, Less Process” (often referred to, in shorthand, as MPLP), a method of processing archival materials more minimally in order to make materials available to researchers and to work through an institution’s backlog of unprocessed collections more efficiently. They advocated a range of strategies, including arranging at a series level (e.g., the correspondence series, rather than sorting the contents of the series into smaller subsets, such as organizing by correspondent or by year), writing only brief descriptions, and relaxing the need to rehouse papers in acid-free folders and to remove metal fasteners. Although archivists continue to debate the impact of the article and the applicability of all of its recommendations, it did cause archivists, as Carl Van Ness’s 2010 response observes, to “analyze why we do certain things in the processing room” and prompted many to rethink their approach.19 Van Ness underlines, however, that Greene and Meissner made certain assumptions about the kinds of resources available to institutions and the reasons for backlogs, which cannot be universally applied. In a study of processing practices and archival ideologies at the Moorland-Spingarn Research Center, Haley Bryant quotes the curator for the manuscripts division who suggested that minimal processing protocols do not do the material justice. The curator argued, “Black materials are undervalued, if not devalued. . . . It just seems like MPLP is undervaluing what we have sought so long to collect and preserve.”20

At the Rose Library, archivists shifted away from some of the more labor-intensive practices, especially as staff curators and processing archivists changed and as the volume of collections coming into the library increased. Processing archivist Sarah Quigley explained in an interview:

Our institutional practice has changed over time as we have embraced backing away from really detailed description. Certainly, in the manuscripts tradition, and in older Rose Library practice, correspondence might be arranged alphabetically by name, so that colleagues, particularly with our Irish poets, would be more obvious. That is incredibly time consuming, and I prefer not to do it that way. . . . There just are so many places where choosing one access point automatically means that you are not choosing others.21

Compared to arranging correspondence alphabetically by name, arranging correspondence by date is faster, and it requires archivists to make fewer decisions through which bias could be introduced, such as deciding between an individual and an organization, or prioritizing one of multiple signatories to a letter.

The way the correspondence series was represented in the finding aid greatly affected the shape of the network graphs we developed in this project. The creators of collections that were described in the older tradition appeared to be highly connected, while collections processed more recently had fewer names in the finding aid, and thus less information about their interpersonal connections represented in the network graphs. For example, in a graph of connections between people, the artist Benny Andrews—who has a collection with an extensive correspondence series arranged and described by correspondent, which was processed in 2006—appears as an exceptionally connected individual. Other collections’ creators may have just as many correspondents, but those connections do not show up in the graph, because they are not named in the finding aid. Collections’ finding aids are composed differently depending on the nature of the collection (its size and contents), the value or priority the library placed on the collection, and, especially, the point in time when the collection was processed. Collections that were processed in the 1990s (the time at which Emory began its collecting focus in African American culture and history) and the early 2000s tend to be much more exhaustively described.

Processing archivists make various judgments when they describe a collection and decide who and what to call out by name. An archivist’s previous knowledge and the priorities of her institution and colleagues will affect what gets named. Archivists who have studied African American religious history might recognize important figures and call out those people in the finding aid, whereas another archivist might not recognize their significance or make the choice to name them in the description of the series. Archivists at Emory highlight connections among collections at their institution, so someone whose papers were also at Emory at the time the finding aid is written is likely to be noted. Archivists also learn about figures and histories in the course of processing one collection that may affect what they recognize as noteworthy in the next one. Finding aids are flexible documents, which is part of their value as tools that can be adapted according to the collection. The biases of archivists, the institution, and the larger society affect what gets named in a finding aid and, in turn, what finding aid data is able to represent. For marginalized, misrepresented, and underrepresented groups including African Americans, such biases can amplify societal inequities, even when an averred value is diversity and access.

For instance, the structure of finding aids and the use of controlled unique resource identifiers in linked open data limit the kind of representation and data we can create from such an initiative. To create the network graphs from the selection of African American collections’ finding aids, Rose Library graduate assistant Ashley Eckhardt and I added tags. Permissible tags in EAD include <persname>, <corpname>, <geoname>, and <title>, which identify people, organizations, places, and titles, respectively. These entities become nodes in the network, linked to either the creator of the collection (e.g., Lucille Clifton) or the collection itself (e.g., Lucille Clifton Papers), depending on where the names appeared within the finding aid. Names that appear in a biographical note or correspondence series are linked to the creator of the collection; names that appear elsewhere are associated with the collection, because a specific relationship to the creator of the collection could not necessarily be inferred as a rule.

When names appear in the finding aids, we added these tags around them, along with a unique resource identifier (URI) for that person or organization from VIAF (Virtual International Authority File), or from Geonames.org for a place name, using Rebecca Sutton Koeser’s Namedropper OxygenXML plugin.22 When James Baldwin appears in a finding aid, he is tagged <persname source=“viaf” authfilenumber=“97644708”>James Baldwin</persname>. The VIAF number enhances the content of the finding aid, allowing us to connect different versions of a name with the same individual and, moreover, increasing the machine-readability of the page. The VIAF number connects the string of letters “James Baldwin” to other resources on the web that reference the same identifier (for instance, his Wikipedia page), which in turn could provide information about important dates, occupation, publications, and so on.

The <persname>, <corpname>, <geogname>, and <title> tags, identified with URIs when available, and the schemas that Koeser and Roke defined in the finding aids database, produce RDF (Resource Description Framework) triples that make statements about the collection in a machine-readable way. Such use of linked open data in archives offers possibilities of connecting collections across institutions, while also incorporating library resources with the structure of the semantic web. An RDF triple is a three-part statement in the form of subject-predicate-object. For example, Louise Thompson Patterson (subject) corresponds with (predicate) Langston Hughes (object). When Elizabeth Catlett is tagged in the “Artist and Subject Files” series of the James A. Porter Papers, a relationship is inferred: the James A. Porter Papers (subject) mentions (predicate) Elizabeth Catlett (object). That triple not only integrates the content of the finding aids with the semantic web but also maps nicely as a network graph, where the subject and object are each nodes and the predicate is the edge that links them.

However, using identifiers from VIAF to enhance finding aids tends to amplify the visibility of prominent, well-established figures. People who have published works tend to have identifiers because VIAF is populated from library catalogs and Wikidata. However, many people listed in this selection of Rose Library collections were not published authors and did not have Wikipedia entries. This technological infrastructure, then, reinforced the preference and bias toward what curator Pellom McDaniels, citing Arnold Rampersad, terms “the big people.” Karen Li-Lun Hwang found something similar in her linked open data project exploring collections related to artist Martin Wong in Metropolitan New York Library Council member institutions. As Hwang assigned URIs to figures associated with Wong, she encountered many people, especially from the Asian American Arts Centre, who did not have an identifier in the Library of Congress, Getty’s Union List of Artist Names, or Wikidata:

The result of this reconciliation process exhibited patterns of inequality in representation and visibility similar to those found in society. Broadly described, those whose contexts intersect more mainstream narratives or stem from the collections of more mainstream institutions disproportionately could find a matching term among these controlled vocabularies, whereas those belonging to more marginalized narratives disproportionately could not.23

Hwang’s observation resonates with our experience at Emory as we enhanced finding aids data with VIAF identifiers.

Controlled subject headings and identifiers also affected the structuring of finding aids at the Rose Library. Emory’s finding aids use metadata standards that impose hierarchical order, including MARC [Machine-Readable Cataloging] and DACS [Describing Archives: A Content Standard] and formats like EAD. For instance, the finding aids must have a “main entry” assigned to the collection, defining the single entity that the collection is about—also referred to as the creator of the collection. This single “main entry” is a crucial part of the logic for linked open data and network data that we were generating from Emory’s finding aids. It is also conceptually clarifying for an archivist processing a collection. If there is a question of how materials should be arranged, the context of that single creator’s life and work can guide the archivist to make a decision.

However, the main entry can minimize the role of other people or organizations that may be represented in a collection. It does not work well, for instance, in the case of families, where one family member has to be identified as the “main entry.” Roke notes the import of the Library of Congress Subject Headings: “If their name is controlled, we will pick them. If there’s a whole bunch of people in a collection, all being equal, we will pick the famous person, or the famous organization, and that’s how we pick.”24 She noted that this practice tends to reproduce systemic underrepresentation and suggests that linked data may offer a path to imagine different structures for finding aids: “Hopefully, with linked open data we can get away from [the focus on one individual creator] because you can then describe a finding aid as belonging to multiple people, and different sections belonging to multiple people, but that would require something fully out of EAD, which is very hierarchical.”25

Although linked open data does offer exciting possibilities for connecting collections across institutions and making more flexible statements about collections than standard library metadata might allow, its promise has limits. Mx. A. Matienzo calls for an examination of “linked data, particularly within the context of cultural heritage, and how it is decidedly not neutral, nor an intrinsic good, but instead as another space in which ideology and systematic oppression are likely to be reproduced.”26 If linked open data can be transformative, that possibility is not inherent to the structure. RDF syntax imposes constraints on how and what relationships can be articulated; in practice, it tends to reproduce social inequities in terms of representation. But there may be ways to intervene. Hwang advocates “the development of open platforms and tools to facilitate the expansion of terms that can be found in the linked open data cloud, for example, the enablement of archives and libraries to contribute their local controlled vocabularies.”27

This network project raised rather than resolved questions of representation in data derived from archival collections. Rather than try to read these networks as representative of cultural and historical networks, I came to read them on a more meta level, as representative of how collections are described and arranged. The schemas we used for much of the finding aid created a relationship between the person, organization, or place mentioned and the collection itself. Reflecting on the project from a distance, I wonder about the decision to attempt to link some of the entities to the creator, rather than the collection, first taken with the more evenly described collections of Northern Irish poetry. If all of the people, places, and organizations were connected to the collection, there would be less of an attempt to read the graphs as representing cultural networks themselves. The data would more clearly be about a collection and its description. Perhaps, then, we could develop methods to read data from finding aids against the grain. Could we read them to identify bias? Could we imagine new structures to describe collections?

If we read the networks for what they tell us about description, we might become aware of what is not there, what we cannot glean from the finding aid. We might become aware of the need to more transparently document the impact and biases of archivists and curators. When archivists arrange and describe a collection, they are not putting things in a predetermined place. Rather, they make choices in what to prioritize, highlight, and name. The finding aid, therefore, is not a direct proxy, but a specific kind of reference to a collection. Archivists and archival theorists recognize that a finding aid is a socially constructed and purposefully incomplete representation of a collection.28 Researchers and those seeking to use finding aids as data should understand the role of the processing archivist and the constraints of the finding aid as a genre.

The question becomes how that social construction and incompleteness of description can be better documented, whether within the finding aid or alongside it. This acknowledgment could become a tool in the larger conversation about who gets to name and describe, as part of a movement to center Black archivists and Black communities in the creation and description of collections. By shifting our point of inquiry here from the cultural networks to the collections’ finding aids, we can interrogate what is made visible and what is lost in current archival description and linked open data practices, and we can move to imagine new structures to access to these histories.