The world wide web, made public in 1991 by Tim Berners-Lee after being developed as an internal tool at the European Organization for Nuclear Research (CERN), has grown exponentially over the past three decades. Since its inception, the web has become a significant site of social, cultural, economic, and political activity. Our lives are increasingly mediated through technology, a current trend which has grown even clearer with widespread remote working and social distancing amidst the COVID-19 pandemic. As one of the co-authors of this article has argued, “without using the web, histories of the 1990s will be incomplete for the most part. Ignoring the web would be like ignoring print culture” [Milligan 2019, 20]. Recognizing the significance of the web to the future historical record, beginning in 1996, the Internet Archive as well as national libraries in Sweden and Australia, began to carry out the widespread preservation of web content. This process, web archiving, can be understood as “any form of deliberate and purposive preserving of web material” [Brügger 2010]. Web archiving has increasingly become part of research agendas for national libraries and archives, as well as memory institutions around the world.

As of writing, the Internet Archive holds over 900 billion URLs and 60 petabytes of unique data (a petabyte being 1,000 terabytes). This figure is probably already dated, as the Internet Archive roughly doubles in size every two years. Despite this sheer amount of data, or perhaps because of it, access to this data has lagged. The sheer amount of data, coupled with the lack of research tools, means that scholars have, in many cases, not been able to carry out fruitful research with this material. Given the importance of web archives to carrying out histories of the 1990s and beyond, this is a serious problem. In other words, the data is there – and considerable expertise has been developed in the collection of web archival data – but the ability to work with it is not. In addition to the problems of scale, there are technical challenges of working with Web ARChive (WARC) files, in which much of this data is saved. We will return to WARC files shortly. Working with web archival data requires an understanding of both high-performance computing and the command line. This is, for the most part, out of reach for many scholars. They do not have the time, the resources, the support, or the training to work with data at scale, meaning that many research questions from the 1990s onwards cannot benefit from web data. In other words, this data is increasingly important for research, but it is too inaccessible for any research at scale.

Our Archives Unleashed Project thus provides scholars and researchers with tools to explore historical internet content and reduce access barriers to large-scale web archival data. Established in 2017 with funding from The Andrew W. Mellon Foundation, the project grows out of a recognition that web archives are critical to understanding the world around us, and that the scholarly community accordingly needs approachable and user-friendly tools to access born-digital cultural heritage. The Archives Unleashed Project exists in several scholarly communities as it is located at the intersection of researchers, tools developers, and libraries. Our goal is to improve scholarly access to web archives through a “multi-pronged strategy involving tool creation, process modeling, and community building - all proceeding concurrently in mutually-reinforcing efforts” [Ruest et al. 2020]. This multi-pronged process manifests itself in three primary ways. First, the Archives Unleashed Toolkit, an Apache Spark libaray for working with web archives directly. Secondly, the Archives Unleashed Cloud, a cloud-hosted infrastructure and web-based front-end to run Toolkit jobs on WARC data [Ruest et al. 2021]. Finally, the project co-hosts Archives Unleashed datathons with local partners. These two-day events bring an interdisciplinary group of people together to collaborate and gain hands-on experience with web archive data.

At the datathons, users are encouraged to use our project tools – they structure the sorts of projects that are undertaken. The Archives Unleashed Toolkit and Cloud were created to provide complementary approaches to working with web archives at scale. While digital content can exist in a variety of formats, both tools specifically work with WARC files, as well as their ARC format predecessor. WARC files, an ISO Standard (28500:2017), are essentially a container-file format that holds collected web resources together. As the web archiving community has standardized around WARCs, this has also made the development of a tool and analytics infrastructure possible.

However, as WARC files are inaccessible to the majority of researchers, so much of the work of the Toolkit and the Cloud revolves around extracting “derivative” files from web archives: familiar formats such as extracted text files, network graphs, or statistical information. For example, when exploring the text of a WARC, several filters can be applied including date, language, keyword, domain, or URL patterns. This also means that projects can be carried out on a wide variety of languages; we have seen successful examples of users working with collections in French and German.

The datathons are the centerpiece of the Archives Unleashed Project’s community engagement strategy. We want to make sure that the Toolkit and the Cloud both reach users and are responsive to their needs. As Niels Brügger has argued, there is great value in “cooperation between web-archiving institutions and Internet research communities” [Brügger 2010]. Our approach to community building has taken several shapes, from providing robust open-source code documentation, running a Slack group with open sign-up for sharing and discussion, regularly blogging, providing a quarterly e-mail newsletter, and – crucially, hosting the datathons discussed here. We also strive to develop educational resources for attendees, building relationships with like-minded projects and institutions (from the Internet Archive to university libraries to national libraries in North America and Europe), and also, participate in scholarly activities that support and foster information sharing.

Our first datathons predated the Archives Unleashed Project. Between March 2016 and June 2017, an earlier project team (including two of this paper’s authors) carried out an initial sequence of four events. These were broader events, primarily focused on networking and building capacity in the web archiving community (described in Milligan et al. [2019]), and included 148 attendees from thirteen countries broadly drawing from web archive curators, digital humanists, and computer scientists.

This paper focuses on the subsequent four datathons as part of our Mellon-funded project. These events were co-hosted with and held at the University of Toronto Library (April 2018), Simon Fraser University Library (November 2019), George Washington University Library (March 2019), and Columbia University Library (hosted online due to COVID-19, March 2020). Collectively, these four datathons have engaged with over 70 participants from seven countries and fifty unique institutions. Participants were predominantly from Canada and the United States, with participants from five additional countries including the United Kingdom, Germany, New Zealand, Egypt, and Hungary. While datathon events were open to a broad spectrum of information professionals, the majority of participants came from the higher education sector: university faculty and graduate students, librarians, and archivists. We had some smaller representation from national archives, non-profit organizations, museums, and independent researchers. In terms of diversity, we did not collect information on racial backgrounds, gender identities, or educational backgrounds; we are accordingly reluctant to make assumptions about our attendees or interviewees.

These datathons had three primary goals. First, they were designed to introduce individuals to tools and methodologies of working with web archives at scale. Secondly, would allow attendees to engage in conversations to facilitate knowledge sharing and scholarly collaborations. Finally, the events aimed to foster community around open-source Archives Unleashed tools and web archive practices. While datathon participants bring a diversity of intellectual and personal perspectives to the events, in general, they can be categorized as access providers, tool builders, and data explorers. For the Archives Unleashed Project, focused on fostering an open-source community, these events would be pivotal for our community engagement strategy.

When thinking of community engagement, we were principally informed by the broad definition advanced by Liz Weaver and colleagues in a paper written for the Canadian Tamarack Institute, an organization dedicated to engaging with citizens to grapple pressing community issues. Weaver et al. define community engagement as “people working collaboratively, through inspired action and learning, to create and realize bold visions for their common future” [Waley et al. 2010]. For organizations and institutions, community engagement is an opportunity to build active relationships with individuals and other entities for mutually beneficial exchanges. Community engagement is vital as it builds a relationship that actively seeks to understand the goals, needs, aspirations, concerns, and values held by a community [Moore et al. 2016]. Without engagement and understanding of a community’s composition, there can be misunderstanding, misrepresentation, miscommunication, and missed opportunities.

Before discussing community engagement models, we need to briefly define what we mean by community. Community is a surprisingly difficult term to define, as meta-reviews of scholarly definitions suggest (in addition to those below, see MacQueen et al. [2001]). We use the definition suggested by Cobigo et al.: “A community is a group of people that interact and support each other, and are bounded by shared experiences or characteristics, a sense of belonging, and often by their physical proximity” [Cobigo et al. 2016, 195]. Indeed, while the web archiving community is largely virtual, our datathons (apart from our one held online due to COVID-19) assembled people together in a physical setting to foster community as well. Accordingly, when we speak of the larger web archiving community, we refer to individuals, organizations, groups, and institutions that have a shared focus on experience and engagement with web archiving activities and research.

Within this broad community there is our Archives Unleashed community, composed of those who who engage with our platforms and support project work. While the web archiving community is multidisciplinary, it can be a resource-intensive process. Accordingly, national libraries and post-secondary educational institutions are overrepresent within the web archiving field (along with the non-profit Internet Archive), providing significant contributions in the form of education, professional development opportunities, services, and tools. The professional backgrounds of these sources are reflected in the demographics and backgrounds of most of our participants.

Community engagement is critical for many other domains, including business and urban development [Fredericks et al. 2016] [Mitra 2016], medical [Joosten et al. 2015], environmental [Fernandez et al. 2016], open-source and technology development [Decker et al. 2015] [Link and Jeske 2017], education [Gribb 1992], libraries [Reid and Howard 2016] [Tharani 2019], and social sciences such as archaeology [Leiuen and Arthure 2016]. Across these studies, however, there are few canonical community engagement models or frameworks; rather, most of the studies reflect on specific case studies or activities, and broad understandings of how engaging communities can benefit a specific group or organization. We did, however, identify two models that could be adapted as a framework to better understand the goals and approaches of the Archives Unleashed Project when it came to community building and engagement. We primarily base our work on the first model, although we draw on elements of the second as well.

The first model is the Open Community Engagement Process (OCEP) model, developed and operationalized by the Water Science Software Institute (WSSI). The OCEP model draws on development methods from Agile Software Development and the Open-Source software development community [Ahalt et al. 2013, 42]. This model approaches community engagement both iteratively and incrementally. Their vision is illustrated as an infinity symbol, with communities continuing to gain knowledge through all steps. Crucially, for our own purposes, the WSSI model features “hackathons,” finding barriers, and disseminating ideas through publications; all of these are critical aspects of the Archives Unleashed Project, especially the first. The limitation of the model, however, is its complexity. OCEP includes fifteen stages, is three dimensional which makes for complicated diagramming, and is difficult to explain. OCEP is also arguably, for our purposes, too focused on software development and does not have the wide range of applications that we want to provide.

The second model that we draw on is the International Association for Public Participation’s (IAP2) “Spectrum of Public Participation.” IAP2’s model describes five critical stages of community engagement (IAP2 2018). The first is to inform, or explain an opportunity available to the community. The second is to consult the community. The third is to involve the community in planning, implementing decisions, project design processes, and ensure widespread understanding. The fourth is to collaborate, or to work together to find solutions. And finally, the spectrum closes with empower, or providing the community with resources and skills to make their own decisions. While community engagement looks different in each sector, the processes are broadly applicable and work well for the Archives Unleashed Project. Understandably, there is no one-size-fits-all-model, as each sector and discipline have unique sets of needs. While we found the stage model of IAP2 attractive for being action-oriented, it lacked the stage and flow organization of OCEP. Accordingly, we combine them as seen in Figure 1.

From the IAP2 model, we drew from the spirit of the five-stages of community engagement: informing, consulting, involving, collaborating, and empowering. In this, we are not alone. The IAP2 model has been applied to “project and program development in fields ranging from health care to environmental planning, particularly in Australia and Canada” [Powell et al. 2010]. As IAP2 focuses on an engagement dynamic originally rooted in government-civic relations, we have adjusted the categories to fit with Archives Unleashed Project priorities, our governance structure, and our activities. For example, our project goals were already defined, so while we do consult with our community for feedback and input on development processes, many overall decisions are made by the project team, not by community vote. This would be the case for many digital humanities projects. It is also important to note, that while we’ve adapted OCEP’s infinity shape for our community engagement framework, stages do not necessarily happen independently or one at a time.

We present the six stages of our model below. Note that the datathon process only appears at stage three of this six-stage process – when we begin to “consult” with the community – but as the first two stages provide invaluable context, we felt it was valuable to provide an overview of all of them.

The first stage is to scope. If we look to the aforementioned OCEP model, we can describe our first stage of community engagement where we scope or identify a problem or challenge, which becomes “the driving question [which] serves as an incentive for a specific subset of the community to participate in the OCEP Open Community Engagement Process] process” [Ahalt et al. 2013, 44]. This stage drives purpose and objectives. Once a problem or challenge has been defined, it is critical to identify the scope of stakeholders that will collectively make up the community we are engaging. In this stage, we ask: who is our target audience, what types of individuals, organizations, or groups are we trying to represent and reach? Who is affected by our “driving question?” With our Archives Unleashed Project, we leveraged previous work on the “Warcbase” project (an Archives Unleashed Toolkit predecessor) to identify needs and barriers of scholars within the digital humanities and social sciences, when working with web archives [Lin et al. 2017]. Through this experience and outreach at diverse conferences and workshops, we gained a sense of who was being affected by high barriers of access to web archives. Scoping and identifying allowed our project to focus on three user-types: academics and scholars (specifically within the humanities, social sciences, and the specific area of the digital humanities); digital access content providers (primarily librarians and archivists), and tool builders (with a focus on those in computer science). Note that this scoping provided the background for the basic structure of our datathons.

The second stage is to inform. The inform stage, as defined by IAP2, is “to provide the public with balanced and objective information to assist them in understanding the problems, alternatives and/or solutions” [IAP2 2018]. Throughout this stage, it was important for the Archives Unleashed team to share information and summaries around the project goals, objectives, and road mapping, as well as bring awareness to web archiving access barriers, and raise an understanding of why our project was important for the community. One of our initial major activities was to develop a presence on several information-sharing platforms. This was to both deliver information and grow a supportive community. Specifically, we used Slack as a way of supportive two-way communication between the team and the community, as well as encouraging peer-to-peer discussions and information sharing practices. An accessible signup form allowed for quick access to our Slack space, and crucially, we could add additional targeted channels for specific aspects of our project, as well as spaces for general discussions. We also set up a quarterly newsletter and regularly-published blog posts.

The third stage is to consult. In our project, this laid the groundwork for the datathons. As IAP2 identified, the purpose of consulting with a community is to open dialogue in which individuals can provide “feedback on analysis, alternatives and/or decisions” [IAP2 2018]. This stage identifies the importance of both asking and listening to voices within the community, and to inform development cycles and approaches by the Archives Unleashed team. We achieved this in several key respects: an advisory board, discussions at datathons (as described in this paper), as well as formally through user surveys and interviews. Indeed, much of the research behind this paper exemplifies our consultation process.

The fourth stage is to involve, which for us, centered around the datathons discussed in this paper. This stage speaks to an active mode of participation from the community. IAP2 defines this stage as a way “to work directly with the public throughout the process to ensure that public concerns and aspirations are consistently understood and considered.”[1] While our project, as an academic one funded by a granting agency, does not involve community participation at the ultimate decision-making level, involvement has taken the shape of working collaboratively with our community on tools development.

Involving our community has been the primary goal of our datathon events, intending to build a community around the tools that would complement and contribute to the broader web archiving scholarly community. As attendees participate in our datathons, they create connections that can support future research collaborations and the sharing of skills and practices with their broader networks. As the OCEP model suggests, by participating in events like our datathons, attendees are exposed to ideas, methods, approaches, skills to address web archiving challenges that may not have “emerged using traditional disciplinary methods and that require synthetic knowledge” [Ahalt et al. 2013, 45]. In other words, the datathon model we adopted as a key activity in our involvement stage draws perspectives and approaches from the community that we otherwise would not have encountered. Notably, involvement overlaps with the consultation stage.

The fifth and penultimate stage is to collaborate and establish partnerships. Here, too, the datathons represented a building of both collaborations with our community and amongst them. The Archives Unleashed Project aims to foster interdisciplinary collaborations as we sit at the intersection of technology, cultural heritage, and researchers. Interdisciplinary and multi-institutional collaboration also provides opportunities for information and resources to be shared more widely. The datathon structure has provided a glimpse into the ways Archives Unleashed has stepped into a role of an intermediary for peer-to-peer collaborations and peer-to-institutional partnerships. As one participant (R4, introduced later in this article) suggested, Archives Unleashed became a “broker of data,” which for some participants created a necessary connection between awareness and knowledge by bridging a gap to accessing web archives data. This has been an active element of our project, and includes formal partnerships with academic institutions to explore their collections, co-host datathons together, and crucially build relationships with other projects and institutions within the field, such as the Internet Archive. Crucially, it has been critical to identify and develop relationships with projects and organizations that could help us foster interoperability between projects (such as data transfer APIs or adapting web-based notebooks).

The final stage, then, is to empower. The IAP2 model defines the empower stage as, “To place final decision-making in the hands of the public” [IAP2 2018]. However, the Archives Unleashed model approaches empowerment in a way which encourages and supports the confidence and skills of an individual to work with web archives, and addresses the objective of lowering barriers to web archives. Beyond the active participation of empowering our users through the datathons, we also seek to give resources to the community through learning guides (text and image-based documents that provide a tutorial-based approach to working with data), videos, and publications which explain our workflows and approaches.

Why the infinity symbol? Ultimately, it reflects our approach to long-term sustainability. After empowering users, we need to consider the scope and think about what the subsequent projects and iterations will look like, and the opportunities to further community engagement and enrichment; we then begin to move through the model again. Cognizant of long-term use of our tools and projects, we are reluctant to post a firm ending to the model.

This was thus our project’s landscape. We wanted to explore the impact of the Archives Unleashed datathon events on community engagement through a robust understanding of these models and the broader scholarly landscape. We return to these two models later in the paper, as we explore how they meaningfully integrate to help make sense of the Archives Unleashed Project’s approach to community engagement.

To understand the impact of our events and inform our proposed community engagement model, we turned to interviews with participants from our event. To do so, we compiled a list of all participants who attended our Mellon-funded datathons between 2017 and 2019 (notably our events in Toronto, Vancouver, and Washington DC). Our primary goal was to craft interview questions and discussions to shed light on steps 3 – 6 (consult, involve, collaborate, and empower). By applying the model to our datathon, we aimed to validate our proposed model.

Our datathons sought to bring together individuals across three discrete categories, as identified by our scoping phase. First, attendees from libraries and archives, – predominantly from post-secondary institutions- or whose professional role was largely related to curating web archives. Secondly, attendees from the tool-building community, or those whose professional roles largely related to developing software for archiving or analysis. And, finally, researchers, or those whose professional role mostly revolved around using web archives for scholarly research. As our datathons are roughly structured to include a third of their attendees from each group, we wanted to interview across these three categories [Milligan et al. 2019]. Some basic information on the flow of the datathon can inform the discussion that follows. Our datathons involve teams working in small groups of between four and six individuals. These groups ideally spanned the three main participant groups (tools builders/library and archives/researchers) and indeed, represented diversity within those sub-categories, spanning the range from libraries, archives, sociology, social media and society scholars, humanists, software developers, government documents librarians, and so forth.

Datathons provided an open and collegial environment that allowed for the organic formation of new collaborations and connections. For some attendees, it was a chance to finally meet people they knew about or follow in the web archiving community but had never had a chance to meet face-to-face with; while for others, it was an opportunity to forge an entirely new network. The datathon also afforded participants the opportunity to meet individuals they would not have met in any other context.

These networking opportunities are especially challenging as there are significant barriers within the web archiving community, especially as it comprises of so many institutions, organizations, projects, and individuals from diverse backgrounds. As the community developed organically over two decades, there is no one overarching body or group to look to, but rather clusters of groups have formed around specific organizations (such as meetings at the Society of American Archivists, or digital humanities conferences, or regional professional groups, or national libraries).

Within the structure of the datathon event, there are various elements and activities geared towards fostering an open and welcoming community, and helping individuals develop a sense of belonging. To encourage team formation, we run a sticky note exercise at the event’s beginning to bring people together into teams. Participants were encouraged to write down research questions on one coloured sticky note, web archive collections of interest on another, and finally, tools and infrastructure on another one. They place their sticky notes on surfaces as organizers physically cluster the notes to identify emerging themes, and encourage physical groups of people around those themed clusters. After three or four iterations, teams would be in place [Milligan et al. 2019]. This method was adapted from the field of participatory design [Walsh et al. 2013].

The sticky note exercise was critical as teams are, of course, the core of the event. Indeed, the teams that emerged out of this were fundamental in fostering group dynamics and belonging. As there were no prescribed research problems or questions, teams were free to decide on the datasets, methods, and research questions – identified during the sticky note exercise and refined while in their groups. Some groups did have a technical expert who they could rely on, but many others would need to work together to approach some of the daunting technical barriers presented by WARC files and difficult tools. Cognizant of limitations of individual’s personal laptops or bandwidth constraints at events, we provide cloud-hosted virtual machines which participants could access remotely.

After working on their projects for two days, the event culminated with final presentations. Teams would have up to five minutes to present their findings, usually through a slide show or dynamic demonstration, and then answer one or two questions from the audience. Judges, selected by the host organization, would then pick one “winner” who received a token prize of Starbucks gift cards (chosen mainly because of their cross-border accessibility).

For this paper, a total of eight semi-structured interviews were conducted via Skype, and subsequently transcribed for reviewing and coding to identify emerging themes and relationships. All interviewees consented to have the interview recorded and agreed to be identified in any ensuing publications with a description of their professional role. While all interviewees had attended at least one of the three Mellon-funded datathons, four of them had also attended pre-Archives Unleashed Project datathon. We will introduce each interview subject again below in prose as they appear in results, but for convenience we use abbreviations for subsequent mentions. They are listed in Table 1.

The interview questions revolved around three general areas. First, we were interested in interviewees’ professional background and experience, as well as their scholarly interests. We wanted to understand their knowledge and experience with web archiving prior to attending the datathon. Secondly, we explored their datathon experience, focusing on their overall experience and impressions from the event, their familiarity with similar events, and crucially whether their scholarly community or practice has changed as a result of the datathon. Finally, we concluded with general questions around their thoughts on the future of web archiving, including challenges, opportunities, and gaps that they deemed relevant to the future of the community. These questions were developed as a means to understand the degree to which Archives Unleashed had successfully created a scholarly community, how the datathons impacted community formation, and what, if any, were the impacts on the broader web archiving community.

The following section is divided into our main themes: backgrounds and pre-existing knowledge of web archiving, the impact of the datathon experience (notably skill-building, exposure to diverse practices, community formation, and inculcation of a sense of belonging), the datathon within the web archiving community context, and finally, general reflections on the future of web archiving and barriers in the field. As noted, these all primarily shed light on steps 4 – 6 (involve, collaborate, and empower) of the community model. For convenience, an overview of findings is provided in Table 2.

The first step in understanding a community has to be understanding its composition. Where do members come from? What technical or social knowledge of web archives did they have? Accordingly, we asked participants about their varied professional backgrounds and scholarly interests. As attendees at a datathon, the participants interviewed all in some way worked with web archives in their current or most recent position: whether they were conducting crawls for institutional collections, providing access to researchers, or creating tools to interact with web archival data. While working in the field, however, when asked about the ways in which they had become familiar with web archiving, many pointed to an accumulation of knowledge based on practice through work experience. Knowledge, in a sense, was a result of trial-and-error and through professional positions, rather than through formal higher education training. This was seen across the board from those with degrees in the LIS, humanities, engineering, and computer science disciplines. Indeed, experience was often obtained outside of the field but applied to it. For example, one librarian working at a large national library (R1) had previously worked with special media formats at scale, an experience which lent itself well to web archiving. Another librarian working at a large research library (R2) had been working with social media data. With both examples, we see a pattern where previous experience with diverse digital formats carried over to their work in web archiving.

All eight participants agreed that the Archives Unleashed datathon had increased their knowledge of web archiving (generally related to higher-level thinking – theoretical aims, goals, processes) outside of any technical skills they learned. One respondent, a researcher at a large U.S. public university (R3), noted that the datathon provided foundational knowledge of some of the working parts and concepts within web archiving, for instance, WARC files, derivatives, as well as the possibilities of web archiving analysis. R2 emphasized that the datathon validated for them the connection between web archiving and scholarly research. For R2, as well as two additional interviewees – one a graduate student and digital humanities developer (R4) and another a digital scholarship librarian at a Canadian university (R5) – the event brought an understanding of how researchers may want to use and look at web archive data. This was important for both tool developers as well as librarians to help inform their ways to support users and colleagues. Finally, another participant – an archivist at a large U.S. private university (R6) – remarked that the event had helped them to think about “what we might change on our end as the staff members creating these collections, to allow for this other kind of research that we hadn't really supported before at our institution.”

With backgrounds established, we wanted to explore the degree to which our events – as an exercise in community engagement – had meaningful impacts on professional practices. We can cluster their responses into four main themes. First, the datathons contribution to skill-building. Secondly, how the event exposed attendees to diverse, interdisciplinary perspectives. Thirdly, how the events fostered community formation; and fourthly – and finally – how our events fostered for some a sense of belonging. We explore each of these in turn.

Despite the varying levels of experience that participants brought to the datathon, all agreed there were several important lessons and skills they took away from their datathon experience. This meant building or adding to their technical repertoires, even amongst those with technical expertise coming in. For those who work with web archive collections as part of their professional practice, such as R3 and R4, their primary focus during their everyday work is to collect information. This means that they rarely have opportunities to, really dig in and work with the data up close. Indeed, this ability to work with the data itself is where the datathon model shined for participants. By digging into the mechanics of data and the intricacies of tools in a small, supportive environment, group collaborations led to discussions around the challenges, limitations, and other structural practices of web archiving. Notably, the wrap-up activity – where the groups presented their final projects – was noted by participants as an opportunity to showcase final projects and to share information.

A number of specific tools were learned, as well. All respondents mentioned the Archives Unleashed Toolkit. Social media analysis tools such as twarc (https://github.com/DocNow/twarc), APIs, and approaches to visualization were also common answers. Notably, R4 noted the importance of learning Spark: “prior to the second workshop, I really had never used Spark before. I kind of knew what it was but just had never spent the time to learn how to use it,” and argued that future work and collaborations would not have emerged if they had not attended the datathon and learned how to use Spark. The overall impact of the event on lowering barriers was also noted by R8, a graduate student at a large private U.S. university, who expressed ongoing efforts to “help archivists to feel more confident or less afraid of the technology of command line and things like that so that they could use Archives Unleashed and the tools that you're developing. Don't be afraid of it. This is great. That's the great thing about the datathon, right?”

Another crucial dimension of the datathon was helping participants make a stronger case for why professions needed to adopt these methods and approaches to web archiving. In other words, it allowed them to better advocate for the value of web archiving within their institutions and communities. As R1, the librarian from a national library described, institutions and organizations are “not as likely to put money towards it or they’re not as likely to just believe in it” without tangible outputs or use cases like one might see coming out of a datathon. This was echoed by R6, an archivist at a large U.S. private university, who indicated that it is hard to advocate for web archiving “when administrators often want statistics and usage numbers and quantitative reasons to support the work, we can’t necessarily say that people are using web archive yet.” As R5, the librarian at a mid-sized Canadian university considered, institutions may not also “realize the commitment that’s needed to do this well.” Through tangible examples, R5 expressed that though their mandate as a digital scholarship librarian, “I find that explaining to them why this is important, how they can use this technology, all of a sudden it becomes less of an abstract fear and more of an oh, I see why this is important and why I want to learn it.” By being able to showcase actual projects from the datathons, participants were able in turn to bring lessons and approaches back to their institutions.

Similarly, participants regularly commented on the benefit of working in small groups (discussed in the previous section) with a diverse range of perspectives unified only in some cases by a shared interest in web archiving. Given the event, this was sufficiently cohesive. Indeed, as R7 (a graduate student and developer at a mid-sized U.S. public university) noted, it represented an “opportunity to network with like-minded people.” The range of perspectives and background groups was in and of itself a real boon. R7 spoke at this at length, highlighting that:

You don’t realize how much you don’t understand until you try to explain something that you think is obvious. So that’s something I think really came out when I was working with teams, was trying to convey things that seemed obvious to me, or intuitive to me, but then other folks were rightly questioning “why would you do it that way?” and I would say, “well how else would you do that?” Basically, going back and forth was really educational.

This back-and-forth would be represented in the final projects as well, as R7 stressed how this creativity was borne out in the final presentations. For example, a researcher with a flair for visualization could make a concept come alive. R7 reflects, in these final presentations, participants could “transform[ing] topics that you never really thought you could take a look at and making them into visual representations.” Similarly, connections between different disciplines could pay off in the long term. R8, a graduate student, was able to connect with an archivist as part of their datathon team, leading to critical insights around data brokering, collections strategies, and beyond; as she expressed, “a lot of things sort of came together just through that one interaction with her. And that led to a lot of insights for me,” noting that they corresponded after the event as well.

While us organizers had the initial goal of building community around the specific Archives Unleashed Tools (i.e., community with our project), coincidently, we were able to encourage connections within the community that did not involve our tools. Indeed, part of this was simply breaking down disciplinary barriers. One of our attendees, R6, an archivist at a large private university, saw the event as helping to “break down the librarian, archivist versus scholar researcher dichotomy ... all of [these] people are in the same room equally as peers.”

Indeed, through the interviews, we saw community formation taking shape as participants emerged as ambassadors within the web archiving community, their home institutions, and professional communities. For example, R3 – a faculty researcher at a large U.S. university – drew a direct relationship between the datathon and their current scholarly practices. For them, the skills and connections learned at the event allowed them to expose colleagues to web archiving practices and demonstrate the possibilities that WARC files provided. Another, R2, the librarian developer at a private U.S. university, noted that it informed their own development work on a related project that used WARCs and some of the web archiving APIs. When it came time to work on this project, the datathon “gave us confidence that when we were trying to do the same thing, that it was the right way to go.”

Within the series of events that were run, participants who attended multiple events matured in their participation as they could mentor new attendees. R7 noted that the first time they attended the event, they were “still learning what a WARC was,” and R3 reflected that “the projects that [they] worked on became increasingly more technical over multiple datathons.” Crucially, these repeat attendees shifted organically into a mentorship role, having the confidence to help other group members based on the technical skills and expertise developed iteratively over datathons.

In general terms, the world of web archiving is unfamiliar and unintuitive to researchers, librarians, archivists, and even some tool builders. At the same time, those who work in the field are quickly confronted with an overwhelming and intimidating amount of data in a relatively idiosyncratic file format (in that WARCs are seldom encountered outside of the web archiving field). To lower these barriers, we want to inculcate a sense of belonging.

A sense of belonging has been defined in a critical article by Hagerty et al. as “the experiences of personal involvement in a system or environment so that persons feel themselves to be an integral part of that system or environment” [Haggerty et al. 1992]. Peer support and a sense of belonging have been identified as two critical factors for overall mental health as well amongst students, and are essential goals of our events as well (drawing on our experience as educators as well) [McBeath et al. 2018]. A sense of belonging is critical to whether an individual participates in a broader community.

The importance of the sticky note exercise in creating a sense of belonging came across in the interviews. As R7 explained, “we did the exercise on the board with the Post-Its and things like that, and I think that really helped me find a group and also find something of interest.”

This was also nurtured through sustained group work. As R3 noted, there was a recognition that the “majority of people were in the same boat because working with web archives is difficult.” By being able to work together, the events for many “normalize[d] the experience with the tools and the technology and limitations” as R3 explained. R4, a graduate student and developer at a large U.S. university, saw the event as akin to a “travelling web archive or reading room,” and by making data accessible to teams (many of whom who could now talk to the curating librarian responsible for the data) we were akin to a “broker of data.” Indeed, these partnerships were mutually beneficial as participants were able to access web archive data, at the same time, attention was brought to unique institutional datasets that exist, but may not be widely known about.

While participants were positive around the overall impact – conveyed through post-event surveys and interviews– ideas did arise around how this event model could either be better refined or perhaps would not fit all of our community-building goals. Some of these, notably the diverse range of technical expertise or disciplines, as well as the slowed momentum of collaboration after the event [Komssi et al. 2015], were expected. Other suggestions led us to engage more deeply with pedagogical literature.

The focus on “data,” implicit in the name “datathon” itself, had some downsides. It led to an emphasis on technical issues and working with discrete datasets themselves, which had the effect of preventing discussion on broader issues to do with policy, ethics, or theory. R3 gave an important example of “issues of representation and what this means and what these gaps mean that aren't data specific” as something that did not feel welcome at the event as a topic of discussion. One could look at a dataset, but this left out questions that might (as R3 again noted) allow us to “look at the whole [web archiving] life cycle.”

With such technical questions and the overall ethos, issues were also raised around the diversity of technical expertise. As teams largely formed themselves through the collaborative team building exercise, some teams ended up without much in the way of technical strength. R6, an archivist at a large U.S. private university, found themselves in such a group and wished that there had been more of an expectation for peers (even if they were in another group) to help “peers who are struggling to get them more towards a middle ground rather than identifying the people who are doing a really stellar job and pushing them farther ahead from the folks who are struggling.” We were inspired by this, and in future grant-funded work, are pursuing a formal mentorship program to help pair expertise with researchers in a more targeted and inclusive way.

Finally, the loss of momentum after the event was a shortcoming identified by interviewees and our project team alike. The short two-day nature of the event means that it largely relies heavily on exploratory projects that may not have a readily apparent route forward to a larger study or publication. It also reflects the difficult nature of the tools, and how outside of the resources provided by the datathon, they may continue to be too difficult. However, we need to consider that engagement is not just a short-term milestone, nor is it just using a particular toolkit or approach. Rather, it is again community. Several participants noted that while they do not use the tools, they do have them tucked away as a resource to recommend when a patron or a colleague might want to work with web archival data.

Participants were all asked about their thoughts on the future of web archiving. While broad in nature, this question was met with optimism from respondents who were generally excited about the direction of web archiving and the opportunities for scholars, access providers, and tool builders (and everyone in between). On the collection side, R2 (librarian developer at a large national library) saw a shift from national to institutional collecting, which would notably involve “large growth in subject-based collecting that is likely to be of more value for both the historical record and for scholarly work.” As well, there was a consensus that web archive collections hold immense value for current and future researchers, and web archives have – as R3 explained – “a potential to add a lot more rigor and also allow us to share data and in lots of other spaces.” We also saw that most participants were happy about the Archives Unleashed Project; R7, for example, noted that they felt “feel very good about the tools that you’ve been developing, like those are exactly the right sorts of tools and tearing down the right sorts of technical barriers that are going to make the adoption happen faster.”

Yet there was also a sense that the process of web archiving analysis might be ahead of where researchers are right now; that we are, perhaps, along with the web archiving field more generally, laying the groundwork for future work. As R2 notes, while there is undoubtedly “long-term massive significance to [this] scholarly activity ... [it will] happen in a timeframe that will be frustrating long ... [but the pace of adoption] should not at all be taken as a reflection on the significance of the work.” R1, a librarian at a national library, remarked that Archives Unleashed “opens the door to (those computational) conversations because it starts to break down [barriers to] the technical part.” Indeed, use of archival collections often lags behind their creation; as R6 argued, “in archives we take a pretty long view of time and history, and just because (few) are using it now doesn’t mean that people aren’t going to want it (web archive data).”

Through this, an emerging community model takes shape for both the Archives Unleashed Project as well as our role within the broader ecosystem; this has implications for other digital humanities projects. It is summarized in Table 3. Archives Unleashed is located within the web archiving ecosystem, with important links to the digital humanities, historical, computational social sciences, and computer science fields. From these interviews and our own experiences, we understand that “if you build it, they will come” mentality does not work, and active engagement to build relationships and communities is essential. The goal through datathons was to build a community around our tools that would complement and contribute to the other connections taking shape.

From our earlier discussion, we identified two community engagement frameworks – the Water Science Software Institute’s Open Community Engagement Process (OCEP) and the International Association for Public Participation’s (IAP2) “Spectrum of Public Participation,” and brought them together in our Archives Unleashed community model. We can see that our combination of the OCEP and IAP2 models works reasonably well, capturing the breadth of activities taking place through our datathons. IAP2 captures the main categories, more or less; and OCEP captures the need to see the cycle as an infinite loop, where we continually navigate and improve as we move through the community engagement model.

Overall, the Archives Unleashed Project has successfully built a community around the scholarly practice of exploring web archives and created unique opportunities for individuals with a wide variance in backgrounds, skills, and experience to connect and collaborate. All participants identified at least one, and in many cases several, tangible skills that they learned. They also spoke to the important aspect of having the opportunity to forge new collaborations, which have positively impacted their professional practices and research collaborations. It also bears out the hackathon literature, which suggests the ability for these events to “enable building a community of users and strategic networks,” [Komssi et al. 2015, 64] seen in our discussions around community formation, a sense of belonging, and skills acquired by participants through the Archives Unleashed datathons.

Without community engagement, projects would live and work in silos. The models we engage with here, as well as the modified version that we advance, are all part of the broader ways in which our projects and others in the digital humanities create a space for and recognize collaborative and interdisciplinary work.