One step up: the importance of failure in a large-scale DH project at the crossroads of disciplines and institutions

In The Importance of Failure, published in 1997, John Unsworth noted that “if an electronic scholarly project can't fail and doesn't produce new ignorance, then it isn't worth a damn.” Unsworth’s essay faced the challenges of balancing planned outcomes with new modes of digital scholarship in a way that was refreshing in DH, pushing for a future where hypotheses are tried and tested and failure recognised as something that takes knowledge forward as much as success. Sadly, his bold call for a reappraisal of what we report did not resonate widely, and we find ourselves still in a landscape where “the trope is one of change, invention, evolution, with overtones of progress and improvement, and with undertones of inevitability and universality”.

In this paper, we revisit Unsworth’s idea in the context of a major DH project that, while delivering our agreed outputs, has experienced some challenges that are singularly and collectively infrastructural, technical, cultural, and epistemological. Echoing Unsworth, we all need to discuss our stumbles and failures on the road so as to responsibly help others move past where we reached; in a time of precarity, slashed budgets, and increasing inequities, we simply cannot afford for a majority of us to keep silent about those parts of our projects which are “trivial, or embarrassing, or simply wrong” (Unsworth 1997).

The authors are investigators in Our Heritage, Our Stories (OHOS), funded by the UK AHRC from 2021-24. This project is a multi-institution, multidisciplinary partnership that includes researchers in digital humanities, archives, history, linguistics, and computer science at the Universities of Glasgow and Manchester; metadata, digital archives, and infrastructure experts at the UK’s National Archives; and a host of collecting organisations and institutions. The aim of the project is to explore challenges in the integration and interconnection of community-generated digital content and collections (CGDC). CGDC is inherently fragile, and classified by the Digital Preservation Coalition as “critically endangered”, but its value cannot be overstated: being community-led, multicultural, and multilingual, CGDC provides an unmatched diversity of materials, peoples, and stories into the UK’s emerging national collection. To do so, the project has taken poorly-described CGDC, enhanced its metadata using an AI pipeline, and made the content more discoverable in order to increase its sustainability and its embedding in research.

OHOS works at a disciplinary crossroads, with different data, specialisms, and institutional perspectives. It is only at this crossroads we find the range of expertise needed to deliver our findings: about the use of AI for otherwise insurmountable distributed infrastructural challenges; about modalities of co-creation and engagement through our post-custodial approach for CGDC; through seeing language in CGDC as a heritage object in its own right; and about key methodological, computational, and infrastructural barriers to linking CGDC beyond the level of entities in order to drive meaningful discovery. This work opens new avenues for further research on the identification, processing and analysis of complex, unstructured data. But the disciplinary crossroads is crowded with contradictions and complexities, and although we have produced all we promised our funders, in working at here we have engaged with many failures around and within our project:

• Infrastructural. The content, tools, and methods that we utilize, as with almost all large scale DH projects, are siloized and challenging to share. OHOS has faced challenges in establishing data workflows that incorporate memory and collecting organisations, academic historians, linguists, quasi-governmental data sources, charities, universities, and an Independent Research Organisation (IRO) in ways that should by now be straightforward, and yet are not. Despite open data being a standard expectation, it is still enormously difficult to separate data from its mode of production for use by open tools and methods, and for some to even conceive of legacy data as independent from its ‘home’ infrastructure.
• Technical. Agile methodologies in play across the heritage sector are completely at odds with the nature of an academic research project, with hugely different expectations around project outcomes, working methods, reporting, and deliverables. Conversely the research community’s understanding of heritage organisations is often outdated and naïve: these are creative industries, with a focus on delivering business-critical and long-term services, rather than in working on short-term discovery.
• Cultural. OHOS is predicated on developing partnerships across universities and IROs, both of which are structurally robust sets of institutions that have policies and processes for conducting and communicating research that are structurally at odds with one another. Research projects expect a certain amount of implicit failure, even when it is not reported, and indeed welcome such failure when it provides alternative pathways or insights; by contrast IROs understandably instead see failure as a lack of responsible planning and the poor management of finite resources.
• Epistemological. How is knowledge created, and what knowledge did we create and why didn’t it work? This raises our key issue: DH is about collections as data, language as heritage, and personal data as primary sources, all of which is challenging to turn into a shared understanding of “deliverables” for IROs who have contractual, constitutional, or statutory limits on how they can engage with new, rich, and complex data.

We discuss these failures, how we have formally and informally addressed them, and what we would do differently in future. OHOS is part of a long journey that will require many technological interventions; these must lay the foundations for ‘infinite’ digital collections that connect archives and records globally (not just nationally), as well as ‘liminal’ digital collections, where archival objects are digitally connected to the communities that hold both objects and expertise needed to bring memory into the archive. Such interventions are hugely important for the future of our field, and will help us transcend digital silos, open up digital archives to new approaches, and help us understand ourselves better using immense sources of rich, distributed, and diverse data. Failures will help us build the way as effectively as successes.