Institutional Perspectives on Digital Collaboration: A View from Down Under
Daniel Russo-Batterham (The University of Melbourne)
Anyone who has heard of the wonderful CRIM project may know by now that members of the ever-growing CRIM community affectionately describe themselves as CRIMinals. Having worked frequently on CRIM over the past years, I am proud to be a member of this group. Although I remain unsure as to the exact nature of our crimes, I would like to test the limits of this moniker with some perspectives from the once penal nation Down Under. Drawing on my experience over the past decade working on Digital Humanities projects across the world, most recently as a Research Data Specialist at the Melbourne Data Analytics Platform (MDAP) in the University of Melbourne, I ask you to join me as I reflect on the landscape of CRIM as I see it currently. I highlight, in particular, areas where partner institutions such as Haverford College, the University of Tours, and my own institution at Melbourne, can meaningfully support these activities.
The purpose of this reflection is not to congratulate ourselves on a job well done—although the collection of essays assembled here suggest some congratulations might be in order—but to identify key aspects of CRIM activities and their institutional contexts that have contributed to success so that we can continue to develop communities of digital practitioners. Like toiling under the hot Australian sun, work in the Digital Humanities is hard, very hard. Digital Humanists interleave specialist domain and technical knowledge in sophisticated ways (perhaps forming a kind of PEn or Fuga, for those keen to get started with the CRIM jargon). When projects of this kind go well, it is important to identify how and why, but also what could be improved. So, here we are, CRIMinals one and all. Let’s use our time in the stocks together to take stock…before heatstroke sets in.
Structure of this essay
CRIM has been a great success in expanding musicological knowledge and bringing an appreciation of the subtleties of Renaissance musical quotation and imitation to new audiences. Crucially, it has also fostered a collaborative community that embraces diversity and inclusion: musicologists conversing freely with Data Scientists in a shared and open language. While it is difficult to isolate the ingredients that have led to the success of CRIM, there are several components present in most successful Digital Humanities projects that are used to structure the following reflection:
1. Building people: Helping people to develop key digital skills is part of every digital project.
2. Building communities of practice: Communities of practice create a sense of belonging and collective purpose that can drive the evolution of digital projects.
3. Building collaborative spaces: Shared physical and digital spaces foster the generation of new ideas and facilitate skill and knowledge transfer.
The above topics are each broad areas of study in their own right, so what follows are some initial reflections that will hopefully lead to more detailed discussions in the future. As a data specialist at the Melbourne Data Analytics Platform, I primarily work on cross-disciplinary research collaborations that have a strong digital component. I am perhaps an outlier in that I have training both in Data Science and musicology, having applied computational methods in my PhD work on seventeenth-century airs de cour. For further information as to my background and the biases that shape this reflection, I refer you to the attached biography.
Building people
Digital projects like CRIM lead to the creation and dissemination of new knowledge and modes of scholarship, at times expediting traditional musicological methods through automation. Yet none of this is possible without a skilled digital workforce. A new DH project may begin with a handful of technical specialists who develop programmatic tools, set up research infrastructure, prepare research data (like MEI files), and so on. Among these outputs are a variety of tools or platforms, such as the CRIM website (crimproject.org), targeted at non-technical domain specialists. Yet for Digital Humanities to continue to grow, perhaps to the point where the term becomes redundant because most research will have substantial digital components, the ability to interact with and think about digital tools must be disseminated more broadly. Helping people to develop key digital skills is an essential part of every DH project. By raising the bar of what are considered foundational digital skills, we can support the adoption of digital best practices, while driving innovation and experimentation that may eventually be incorporated into these best practices.
It is tempting to divide the CRIM community into tool creators and tool users and at times this bifurcation is useful for defining roles and responsibilities within the broader community, but in reality this is much more of a continuum. Tool creators are also tool users, and tool users can develop sufficient technical ability to contribute directly to the advancement of the tools. Even before developing any sort of technical proficiency, users contribute to the refinement of technical tools with feedback on what works well and what does not, shaping debates on where future work should be focussed. For another perspective on this continuum, I would invite you to refer to the excellent presentation that Frans Wiering gave in Tours, France 2022 on Computational / Digital Musicology.
Problem solving and skill acquisition at in-person events
One inescapable feature of living and working in an Australian university, particularly in Melbourne, is that the majority of potential international collaborators are far removed, both in distance and, often, timezone. More broadly, the reliance of DH teams on diverse expertise that is seldom co-located means that collaboration via Zoom, Slack, and so on, is the norm. We can rightly laud these technologies for allowing us to continue to work through lockdowns (of which Melbourne had plenty!) and across international borders, but digital work is most effectively conducted when people can interact in person. This is particularly true when it comes to problem solving and skill acquisition.
At CRIM@Tours and CRIM@Haverford events, in June and October 2022 respectively, participants from around the world gathered for several days of intensive work with CRIM tools and notebooks. Although there were welcome and valuable contributions from participants online, these events were primarily in-person. These days comprised a mixture of presentations, where CRIMinals would demonstrate their findings (and grievances) using various CRIM tools, and a series of workshops where small groups would focus on particular aspects of CRIM, often led by an expert in a particular area, such as short introductory Python and Pandas workshops for those still getting comfortable with some of the technologies that underpin CRIM. One of the great strengths of the CRIM conferences, workshops and other activities is that they are inherently applied. Participants are asked not just to reflect on the computational reports produced by various CRIM python packages, but to actively manipulate the output, and examine the strengths and weaknesses of the computational methodology. The diversity of backgrounds of CRIM attendees helped to break down boundaries that can form along the lines of language spoken (see Granjean 2016) or research domain.
Preceding these conferences, many online zoom sessions had been held to troubleshoot various aspects of CRIM. Importantly, however, people reported during the in-person workshops in Tours and Haverford that complex programming and technical concepts were much easier to grasp face-to-face. While in-person learning is not always practical for a community as diverse and geographically dispersed as CRIMinals find themselves, the success of these events shows that when opportunities to gather present themselves, there should be a particular emphasis on skill acquisition, skill transfer and other types of knowledge exchange that are easier face-to-face. Richard Freedman and other technical leads have played a key role in teaching core computational concepts and tools, and how they might be applied to this repertory, but collaboration with these tools in cross-disciplinary groups facilitates peer-to-peer learning.
It was also clear that very few participating researchers at Haverford and Tours had received significant and sustained Digital Humanities training from their respective institutions. To some extent this is OK, since learning through collaboration is a defining feature of the kinds of communities of practice (discussed further down) that we are looking to build. Nonetheless, the greater the technical foundations across a given group, the more scope there is for innovation at meetings of this kind. Given this, we might reflect briefly on what institutions are doing to support digital and data skills. This is a huge topic, but a few current examples of where institutions are looking to support data-intensive research should nonetheless be useful.
Institutional training
As the value of digital and data skills continues to grow, research institutions are investing in training staff and students in technical skills that support innovation. While there is still too little training and support for Humanities research, it is worth exploring some of the different models for upskilling researchers and how they might relate to projects like CRIM. The brief examples here are drawn from institutions I have encountered directly such as The University of Melbourne, The University of Tours, Haverford College, The University of Pennsylvania, and the University of Toronto, but they represent modes of digital skills pedagogy that have been used with varying degrees of success in other parts of the world.
Institutional skills training and interdisciplinary research collaboration
At the Melbourne Data Analytics Platform, we provide skills training and knowledge transfer through collaboration on projects that span many disciplines. Every year, small, specialised teams within MDAP work with one or more researchers from around the university, leveraging digital methods to uplift research projects (see https://mdap.unimelb.edu.au/previous-collaborations/). The limitation of this “collaboration model” is that it does not scale directly to a large number of researchers. Many of those we work with, however, go on to train their colleagues, who can then train others in digital methods. Moreover, both MDAP and other university groups are looking to complement these collaborative partnerships with more conventional classroom-style training, such as introductory Python or R.
Internships, Studentships, and Mid-Doctoral Fellowships
Many students at the University of Melbourne get their first taste of Digital Humanities and digital methods more broadly through one of a number of internship programs, a pattern I’ve seen successfully employed elsewhere, most notably the studentships at Haverford College. While MDAP routinely involves student interns on research projects, they are only occasionally coming from a Humanities background. It is important then that each year around ten postgraduate student interns are paired with academics in the Faculty of Arts, through the Digital Studio, a “collaborative research hub” that “aims to humanize and leverage digital technologies to transform the ways in which teaching, research and scholarship shape the social, ethical and cultural dimensions of our world”. After attending half a dozen training sessions on digital methods, from data wrangling through to data visualisation, interns work on their allocated project for about 45 hours, spread across several months. This is a modest time commitment from the students, to the extent that you could question the effectiveness of these limited internships in imparting genuine skills and knowledge. Programming to a sufficiently high standard to produce poignant research outputs, traditional or otherwise, can take years of study. While students can and do gain some real skills during this timeframe, by far the biggest benefit students get is gaining some initial exposure to the world of Digital Humanities and digital tools. Many then continue their study of programming languages and build authentic and unique digital expertise by applying these methods to their PhDs and postdoctoral work.
A far more extensive form of applied digital skills training can be found at the University of Pennsylvania’s Price Lab for Digital Humanities. Here, a number of PhD students are offered mid-doctoral fellowships, again building on the idea that students can use these skills through their theses and throughout their careers. While somewhat unique, this is certainly a model that could be explored elsewhere. As a large Digital Humanities project, CRIM itself is an initiative that supports not just established academics, but students at the start of their careers. Many major improvements to CRIM can be attributed to Haverford students. It has been a pleasure to provide some supervision to students both at Melbourne and Haverford (with varying degrees of formality), and I have arguably learnt as much from them as they have from me.
Academic Specialists and the “Third Space”
It is not a coincidence that current initiatives to uplift data-driven research are focussed on the next generation of researchers, so-called “Digital Natives”. Yet when careers in academia are still judged largely on traditional publications, it is imperative that institutions provide appealing career pathways for those focussed on non-traditional research outputs, like all the digital artifacts that comprise much of CRIM itself. Developing research software is an incredibly challenging task and it is vital that these skills are recognised accordingly. I noted earlier that technical tool creators and tool users are far from mutually exclusive groups, and yet the former are likely to have dramatically different career trajectories and expectations. In most instances, programming, software development, and other technical competencies are classified as professional skills, rather than representative of an academic. But in practice, the line between academic and professional work is increasingly indistinct. Whitechurch notes, for instance, that “At the same time as professional staff are acquiring academic credentials, some academic staff are moving in a more project-oriented direction. This has effectively created a Third Space between professional and academic spheres in which lateral interactions, involving teams and networks, occur in parallel with formal institutional structures and processes, and give rise to new forms of management and leadership” (Whitchurch 2012). Toward this end, the University of Melbourne, alongside teaching and research academic staff classification, has a broad category called “Academic Specialist”, which policy documents describe simply as “positions which are appropriately classified as academic but which are not expected directly to undertake teaching or research activities”. Work to better define this role is ongoing and I suspect this definition may change over time.
To appropriately reward digital specialists and prevent them from leaving for industry en masse (where they can often demand a significant salary increase) we need to recognise different modes of scholarship and place them on equal footing to traditional research outputs. The San Francisco Declaration on Research Assessment (or DORA) goes some way toward these objectives, but is just the start of what must be a wholesale shift in the way we evaluate academic work and academic research. While the document is heavily focused on scientific perspectives on research, which is a major limitation, the general principles can be applied to other areas of research. DORA advocates for a more holistic evaluation of research and different kinds of contributions, such as web applications, software or datasets. CRIM and similar projects have a crucial role in providing valuable learning opportunities for academics at all levels, while demonstrating to the broader scholarly community that expertise in digital methods must be valued and recognised accordingly.
Communities of practice
Communities of practice create a sense of belonging and collective purpose that can drive the evolution of digital projects. Although there has been significant research into such communities, it is rare to find a collective of researchers that lives up to the name, since the core “practice” component is so often missing. In this case, what is left is perhaps best described as a “Community of Interest”. Practitioners of the Digital Humanities, in musicology or any other domain, need to develop hands-on experience working with digital methods. Since researchers seldom develop technical competencies on their own (i.e. self-taught), large projects like CRIM can foster the development of whole cohorts at once, while institutions also play a role by ensuring that training, opportunities for practice, and spaces that support collaboration are readily available. Together we can learn not just how to code, but how to navigate and ask questions of research data in an environment where computational tools can automate certain tasks, while leaving some of the more nuanced thinking to us humans. There are varying definitions for a community of practice (see the comparison in Cox 2005), but for now we can usefully adopt the following provided by Wenger:
Communities of practice are formed by people who engage in a process of collective learning in a shared domain of human endeavor: a tribe learning to survive, a band of artists seeking new forms of expression, a group of engineers working on similar problems, a clique of pupils defining their identity in the school, a network of surgeons exploring novel techniques, a gathering of first-time managers helping each other cope. In a nutshell: Communities of practice are groups of people who share a concern or a passion for something they do and learn how to do it better as they interact regularly (Wenger 2011: 1).
In the case of CRIM, the “shared domain of human endeavor” is the Renaissance of Mass, but also computational or digital musicology more broadly, since we have seen CRIM tools ably adapted to answer questions of other repertories.
Developing shared language and conceptual frameworks
One of the most striking features of the CRIM community is the development of a shared vocabulary for describing the Imitation Mass repertory. Someone new to this area of study might be overwhelmed by the speed at which terms like soggetti, PEns, fugas, imitative duo, modules and so on fly around the room. Whether to apply one label or another can, and frequently has, been cause for a polite, yet surprisingly intense debate. Every subdomain of musicology has a shared set of vocabulary and theoretical foundations that people both know and care about enough to argue over.
More unique to the CRIM community is the adoption of a set of nomenclature more closely tied to computational thinking than to any musical corpus or theory. Through gatherings online, at Tours, and at Haverford, we have become accustomed to mentions of ngrams, heatmaps, Levenshtein distance, tabular data, machine learning, regular expressions, false positives, network diagrams, Jupyter Notebooks and much more. This shared language underpins collaborative digital work, but it was not something that developed easily. It is fair to say that there was some broad reluctance toward sullying otherwise pure musicological discussions with computer speak. Yet as CRIMinals became more familiar with the tools the technical team had developed, the adoption of this shared technical language became increasingly natural.
Familiarity with this language and the underlying concepts also had a significant impact on the methodological and conceptual frameworks applied to the CRIM repertory. Many have quite rightly used CRIM’s suite of python tools to do what might otherwise be done by hand and do it faster. This is certainly an excellent use of the technology. Using python to quickly identify interesting patterns in an automated manner can free up time for a qualitative evaluation of these passages in the context of the score. This, already, is a great step forward because it demonstrates that programmatic tools do not need to replace “traditional” musicological work, but can instead complement it. Recent additions to CRIM Intervals make this even easier by showing relevant parts of a score directly in a Jupyter Notebook. We have also seen CRIM tools used to do musicological analyses that would not be practicable without the aid of computers, ranging from large-scale frequency analysis of different features, to complex network diagrams that plot relationships among works, to using the tools as a way to reduce bias (Margot and Schubert 2023). Germane to this creative use of the technology is an understanding that the many possible abstractions of a corpus that computational tools can produce—a list of ngrams, for instance—can be meaningfully related back to the music itself and our experience of it.
CRIMinals are so diverse in their backgrounds and skills that they often belong to multiple interconnected research communities to their collective benefit. A good example is the enormous global community that has built up around the python programming language. Most of the CRIM website and CRIM analysis tools are built using python. This is an excellent choice that will allow new digital humanists to explore the landscape freely. As a free, open source, general-purpose programming language, python can be used to complete many routine tasks in the Digital Humanities. It has a huge ecosystem of existing packages that can support complex tasks required when representing and manipulating music as data. Yet the most impressive “feature” of the language is the great community that can answer questions.
Getting large institutions to collaborate
The CRIM community brings together researchers from many different institutions. This is quite a rare accomplishment. While many institutions foster the development of communities of practice within their walls, such communities are at risk of becoming silos. With around 50,000 students, the University of Melbourne is one of the larger education institutions in Australia. My personal experience is that larger institutions find it more difficult to collaborate with other universities quickly and effectively. There are several reasons for this. One is that significant time and effort goes into collaborating across faculties within the university, which typically takes precedence over external engagement. Second, smaller universities, such as regional universities in Australia or some of the smaller Liberal Arts colleges in the US, are often obligated to collaborate closely with other institutions, since by pooling resources and expertise, they can offer a broader range of subjects or work on a broader range of research projects. This is evident, for instance, in the close relationship between Haverford College and Bryn Mawr College, both of which a part of the Liberal Arts Collaborative for Digital Innovation as well as a Tri College Consortium. The apparent difficulty with which large universities collaborate is problematic when it comes to the development of projects like CRIM that require more expertise than any one institution, even a large one like Melbourne, is able to provide. Thankfully, there are Australian collectives looking to facilitate digital partnerships across institutions, such as the Federation for the Advancement of Victorian eResearch.
Building collaborative spaces
Shared physical spaces help to generate new ideas
When reflecting on Digital Humanities projects like CRIM it is easy to overlook some of the practical factors that can help interdisciplinary work succeed. Universities and other research institutions have an important role to play in providing physical spaces that support the free exchange of ideas and the opportunity to mingle with those outside an individual research unit. As an extreme example, the University of Melbourne recently built a precinct called Melbourne Connect, which it markets as a ‘Nexus of Innovation’. The website modestly claims that by ‘Bringing together world-class researchers, government, industry, SMEs, startups, higher-degree students, artists, and Science Gallery Melbourne, Melbourne Connect is a purpose-built innovation precinct right in the heart of the Melbourne Biomedical precinct’. While I won’t delve into the theory of what makes a space good for collaboration or the different forms these spaces may take (see Montanari, et al 2020: 3-5), I can say after working in Melbourne Connect for the past year that the collocation of people from diverse backgrounds, however contrived, does open new possibilities, albeit with some quirks and caveats. Open plan offices, for instance, may facilitate collaboration but they can also make deep work difficult. Similarly, a proliferation of glass walls provides a sense of spaciousness and connectedness but limits privacy. The study of patterns of similarity and imitation in Renaissance masses and models seems far removed from many other fields, but our representations of music as sequences of intervals, has many parallels in fields from Bioinformatics through to Linguistics that we can use for inspiration. Spaces that cater to serendipitous encounters are vital to innovation. I’m sure attendees at the October Haverford conference would agree that the Jaharis Recital Hall was an active participant in proceedings, providing an aesthetic and acoustic space that significantly enhanced the discussion and performance of Renaissance music.
Shared research infrastructure and recipes
While writing or at least tweaking code is within the reach of most researchers, getting started still presents significant barriers. It would be difficult to overstate the impact that such practical matters can have on the establishment and subsequent growth of communities of practice. For CRIM, the availability of JupyterHub, a shared online environment for computational work, greatly facilitated this process by providing a working analytical environment to researchers with minimal setup. The deployment and maintenance of this research infrastructure often falls to universities or other research institutions. From the perspective of the researcher, they can simply log in, without having to worry about where or how their code is running. There have been recent efforts in Australia, through groups like the Australian Research Data Commons, to provide broad access to similar platforms across many institutions, but research infrastructure for the humanities poses unique challenges and continues to lag other domains.
Within JupyterHub, a set of prebuilt (but constantly evolving) Jupyter Notebooks provide recipes for using CRIM tools to perform routine tasks. Richard Freedman and other contributors deserve a lot of credit here for putting together useful sequences of code snippets in the code cells that make up much of a notebook, augmenting these with comprehensive text descriptions in the accompanying markdown cells (https://github.com/RichardFreedman/CRIM_JHUB). While CRIM Intervals provides a set of relatively simple functions and data structures to the analyst, it has quickly emerged that most researchers new to this work are more comfortable running and adjusting an existing notebook “recipe” than starting from scratch. This will, of course, change over time as the CRIM community grows in experience. The reliance on these recipes risks creating a certain sameness across analyses. Yet the success of these recipe notebooks and the JuptyerHub digital space in helping people to start using computational tools is nothing short of astounding, demonstrating the value of reducing practical barriers to entry wherever possible.
Conclusion
The CRIM project has been a world leader in the development of innovative digital approaches to the Renaissance Imitation Mass. Moreover, it is a valuable case study in how institutions can support large Digital Humanities initiatives through the provision of digital skills training, by providing career pathways for digital specialists, by fostering the development of communities of practices, and by building and maintaining physical and digital collaborative spaces that support research.
Bibliography
Cox, Andrew. 2005. “What Are Communities of Practice? A Comparative Review of Four Seminal Works.” Journal of Information Science 31, no. 6: 527–40. https://doi.org/10.1177/0165551505057016.
Grandjean, Martin. 2016. “A Social Network Analysis of Twitter: Mapping the Digital Humanities Community.” Cogent Arts & Humanities 3, no. 1: 1171458.
Montanari, Fabrizio, Elisa Mattarelli, and Anna Chiara Scapolan, eds. Collaborative Spaces at Work: Innovation, Creativity and Relations. London: Routledge, 2020. https://doi.org/10.4324/9780429329425.
Schubert, Peter, and Sylvain Margot. 2023. “With Baccusi in the Jacuzzi; or, How I Learned to Stop Worrying and Love Numbers.” In CRIM Project Perspectives
Essays and Experiments from Citations: The Renaissance Imitation Mass, edited by Richard Freedman. Read Online
Wenger, Etienne. 2011. “Communities of Practice: A Brief Introduction.” Read Online
Whitchurch, C. 2012. Reconstructing Identities in Higher Education: The rise of ‘Third Space’ professionals. London: Routledge. https://doi.org/10.4324/9780203098301