Similar to many projects now associated with digital humanities, nineteenth-century philologists confronted a growing evidentiary problem: the scale, dispersion, and uneven quality of textual material exceeded the capacity of existing scholarly methods. Lexicography, in particular, lacked reliable and systematic sources from which linguistic evidence could be drawn. Among the various responses to this challenge, this paper argues that the Oxford English Dictionary (OED) stands out as a sustained effort to resolve an evidentiary mismatch through infrastructure development rather than interpretive innovation. Its editors confronted problems of excess, heterogeneity, uncertainty, and fragmentation long before the emergence of the language of “big data” or digital infrastructure. This paper reframes the OED as an infrastructural response grounded in coordination, governance, and workflow design (Edwards et al. 2013; Borgman et al. 2020; Mayernik et al. 2025).

This paper offers a historical case study read through the lens of the data problem, defined as the organizational and epistemic challenges that arise when available evidence exceeds the capacity of existing scholarly methods and infrastructures to support shared inquiry. The analytic concepts deployed here are iteratively developed through comparative analysis of a purposively selected set of humanities infrastructure development cases (n = 10). Using infrastructural analysis, the paper reconstructs and examines the emergence of the OED not as an early instance of digital or computational practice but as a historically specific response to a data problem. Contemporary discussions often treat data problems as recent disruptions tied to computation and digital tools, obscuring a longer history of large-scale humanities work concerned with the organization and coordination of evidence. The OED demonstrates that such problems predate digital technology and have long been addressed through infrastructural design.

The paper reconstructs the OED’s development as durable research infrastructure by examining editorial organization, distributed volunteer labor, citation slip systems, and the network of scholarly societies that supported the project, most notably the Early English Text Society. Using this case, the paper introduces infrastructural agency as an analytic concept, locating agency not primarily in individual acts of interpretation but in durable systems that organized evidence, enforced standards, and sustained collaboration over decades. This agency extended beyond lexicography into publication practices through shared editorial and typographic protocols at Oxford University Press, including Hart’s Rules, which functioned as standards for large-scale textual production. The central challenge facing the OED was not the absence of evidence but the management of scope, heterogeneity, and coordination. The paper shows how the OED enables a comparative history of humanities infrastructure while reframing contemporary debates about infrastructure development away from abstract models and toward historically grounded infrastructural practice. Reconstructing this history clarifies how digital scholarship in the humanities might be approached today by emphasizing infrastructure as a problem of alignment, governance, and durability rather than computation alone.

« the published texts of a project do not tell the same story as participant observer story. » (Adam 1998)

En s'intéressant aux coulisses de projets informatiques, notamment en IA, Adam souligne l'écart entre le discours officiel d'une science et la pratique concrète : plus qu'un simple décalage, c'est une distorsion qui, en effaçant par différents mécanismes la matière concrète et participative des conceptions, permet d'ériger une image propre et normative d'une technologie. Lisser les coulisses n'est pas propre aux sciences dures : les sciences humaines invisibilisent également des dimensions collectives ou triviales pour proposer des récits abstraits de l'émergence de la connaissance (Waquet 2022).

Pour les HN, on peut s'interroger à savoir si les présentations et documentations des projets n'ont pas hérité du même réflexe à blanchir et nettoyer l'histoire de l'établissement d'expertises qui ont, nécessairement et dès les débuts de ce champ disciplinaire, été le fruit du collectif et d'un certain artisanat (Index Thomisticus, Nyhan & Terras). Au-delà de l'invisibilisation d'acteurs passés ou actuels (« petites mains », Mellet; « travailleurs du clic », Casilli), ce papier souhaite explorer l'idée de pratiques grises comme l'ensemble des connaissances tacites -- know-how -- qui échappent aux littératures blanches et manuels promettant une application sans accroc. Cet ensemble se construit d'itérations, de non-ordonné, d'errance et de social, soit tout ce qui est évacué par la postérité pour définir un domaine d'expertise, donc de certitudes : les discours informatiques gomment ainsi les traces de manipulations et de pérégrinations concrètes (tableaux blancs, stylos de couleur, café) pour déterminer une science dure.

Les cultures numériques ont favorisé l'émergence de pratiques remettant en question des notions traditionnelles comme l'originalité, menant les écritures modernes à opérer par rapiéçage (Goldsmith 2011) ou patchwork à partir des bribes copier-collées d'ailleurs. C'est en ce sens que le terme de bricolage permet de définir les pratiques improvisant de nouvelles littéracies, où la pensée s'élabore dans l'action (Ingold 2013). Éloignée des imaginaires mécanistes, montrant la technologie comme froide, rigide, assurée, procédurale, cette proposition explore la notion de pratiques grises pour tisser une autre histoire scientifique et penser le bricolage comme une forme d'intelligence par le fil. À la réflexion d'Adam s'accordent plusieurs travaux s'intéressant justement au tissage comme technologie de pointe à part entière (Tripaldi 2022), sous-estimée parce qu'historiquement liée aux femmes.

Cet imaginaire tissé est loin d'être étranger à la constitution d'une culture numérique : le métier à tisser Jacquard de 1801, considéré comme la première machine programmable; les projets comme E.A.T. fait de fils, de collaborations et d'expérimentations plus ou moins orchestrées (parce que le lieu de répétition et celui de représentation étaient différents, Mellet 2025); jusqu'au réseau de données actuel. Or dans la conception de ces maillages, comme ceux des systèmes informatiques, sont inscrits implicitement des concepts antinomiques (féminin/masculin; noble/ignoble; traduisible/intraduisible). Inspirée de l'archéologie des médias, la réflexion sur les pratiques grises sera étayée d'une étude de coulisses de projets en HN partagés (sur des repositories git). Cette conférence sera donnée en langage switchable, soit en alternant l'anglais et le français.

The digital humanities – as a heterogeneous collection of methods, pedagogies, and disciplinary formations – is no stranger to debates over the approaches contained under its umbrella. Ethnographic approaches to the digital, however, are often conspicuously absent from these discussions (with isolated exceptions such as Clement 2016 and Hsu 2014). In this talk, I outline the history and methods of ethnographic approaches to the digital, focusing on digital ethnography, multimodal ethnography, and ethnographic science and technology studies. Focusing particularly on digital ethnography, which takes digitality not only as its object of study but also its method, I argue that the ongoing divide between ethnography and the digital humanities is the result of historical disciplinary boundaries more than any concrete methodological or philosophical differences. The sparse commentary that does exist on this divide typically attributes it to ethnography’s own predilection for studies about rather than with technology (e.g., Varis 2020), a perspective which fails to account for a wide range of historical and contemporary ethnographic research practices. Ethnographers have already been moving towards DH – from ethnographic studies of DH (Bailey 2015) to growing calls for ethnographers to embrace DH methods (Fraser & Linares 2022) – and it is time for digital humanists to, in turn, move towards ethnography.

As a method primarily concerned with embodied, performative, and ephemeral cultural phenomena, ethnography at its core is a robust framework for the study of the untranslatable; (digital) ethnography offers a robust method for generative engagements with the untranslatable (Lebner 2022), the opaque (Porfírio 2022), the incommensurable (Chambers-Letson 2018), and the unknowable (Besteman 2020). As such, and in line with approaches in feminist data studies (D’Ignazio & Klein 2020), ethnographic approaches can allow for a reinterpretation of the meaning of knowledge production itself within DH. What would it mean to view our data as agential interlocutors? How might ethnographic approaches to fictionalization generatively intersect with quantitative, data-driven DH analyses? How can data visualization methods in DH be enhanced through the incorporation of techniques from multimodal, visual, and graphic ethnography? How can reflexive approaches born from feminist and postcolonial ethnography help clarify ongoing conversations in DH around data colonialism, digital surveillance, and the neoliberalization of the university?

This talk draws on my experience conducting ethnographic research in virtual worlds and mixed reality environments, as well as teaching both DH and ethnographic methods to undergraduates, to demonstrate how ethnography can supplement not only our DH methodological toolbox but also help transform the way we teach DH in our classrooms.