The digital and computational

The affordances of ‘the digital’ are well articulated (Naughton 2014). One is the ability to bring together objects that could not come together physically, in digital juxtaposition or superimposition. Democratised access is an affordance in terms of access to content and also in terms of analysis, illustrated for example by crowdsourcing that can engage volunteers at scale. Most pervasively we have the common medium of digital data – or the bitstream – to represent all forms of cultural artefact, and blurring a boundary between born-digital and remediated content.

An immediate critical note is needed: we are not asserting that the digital representation of an object is a replacement for its material form – for example, there is typically a loss of information associated with any selecting or encoding, of content or metadata. Furthermore, no infrastructure is agnostic, and this is as true in the digital as the analogue sphere. A common implicit presumption that ‘digital’ is naturally more authoritative or complete must be challenged. In other words, disaffordances of the digital can be identified too, and this balance is surely necessary in any discussion about failure. Furthermore, affordances have consequences: for example, while it is an affordance that digital content is significantly easier to copy and distribute than material goods, this also means licensing, rights and business models need to evolve.

Large-scale analysis and automation are well known digital affordances – in fact computational ones: the digital data opens the door to today’s computational approaches and prepares us for tomorrow’s. Less widespread perhaps is simulation, which enables us to explore ‘what if’ scenarios in our research inquiry; indeed it lets us experiment and fail safely. We might look at this affordance as close reading by digital prototyping, a kind of ‘experimental’ humanities (De Roure and Willcox 2017). We see adoption of simulation in augmented and virtual reality as a humanities method too, for example in historical reconstruction of buildings and 3D modelling of artefacts. Because these digital artefacts are easy to construct, we can readily try out alternatives in our process of enquiry – and in fact this process can be supported by automation.

Programming

All these affordances have digital content in common, and they are enabled by software tools. What is important about these tools is that we are able to create them, to adopt and to adapt them. This is a powerful evolution in our research infrastructure, and prompts the suggestion that programming is a crucial affordance of the digital too. No infrastructure is agnostic, so available software tools are a lens which might lock us into particular modes of analysis – we must not forget this, and knowing it will help us address it because software is a particularly versatile and adaptable instrument.

Software can itself be the subject of critical discourse, and its evidence in the archive – even if it no longer works – is evidence of our understanding, aims and methods. There is excellent scholarship around early software (Berry and Marino 2024) and what is now termed ‘Digital Humanities’ has a history as ‘Humanities Computing’, or even non-numerical computing (Michie 1963).

However, software can be hazardous too. Unsworth articulates the role of software and eloquently discusses the ‘software waste cycle’ (Unsworth 2020):

Software developed by individual researchers and labs is often experimental and hard to get, hard to install, under-documented and sometimes unreliable. Above all most of this software is incompatible. As a result it is not at all uncommon for researchers to develop tailor-made systems that replicate much of the functionality of other systems, and in turn create programs that cannot be reused by others and so on in an endless software waste cycle.

This is a disaffordance but we can do something about it. The Software Sustainability Institute (SSI) was founded in the UK in 2010, dedicated to improving software in research across all disciplines – indeed to address this waste cycle, helping to ensure benefit from funder investment in software creation.¹ As well as providing training, SSI introduced the notion of the ‘Research Software Engineer’ (RSE): these are skilled colleagues who contribute to research by developing software and engaging with the problem. The RSE role has enjoyed significant uptake (Cohen et al. 2021) and this spread extends to arts too (Ma et al. 2024).

Software and skills

SSI undertook a large-scale survey of the UK arts and humanities research community to establish practices and skills in the use of digital tools and software (Taylor et al. 2022). The report highlights diversity of practice, evolving interdisciplinarity and collaboration, a wide spectrum of engagement and the importance of communities of practice. Recommendations also identify skills and knowledge development, data management and sustainability, and supporting careers including early career researchers and research technical professionals.

Indeed, supporting communities of practice in software creation, documentation and training may be a particularly effective intervention in the humanities context: it is a means of sharing practice from the ‘early adopters’ to build capacity. A strong example of this approach is the consortia of Huma-Num, the French Research Infrastructure for the Social Sciences and Humanities, where collective consultation defines communities and their technical resources.²

SSI itself has sustained for fourteen years to date, helped by its ‘collaborate, not compete’ ethos and wide disciplinary breadth. Its current work focuses on four of the software sustainability priorities we perceive today: capable research communities, widespread adoption of research software best practice, evidence-driven research software policy and guidance, and broadened access and contributions to the research software community.³ In this way, SSI is building resilience against failure into the research community and addressing the software waste cycle.

The new primitives

While digital computational resources have grown over recent decades, so too has the engagement of citizens with the digital world – we were once a community of scholars with access to a small number of computers; now we operate in a world of crowd and cloud, and a new abundance of both data and computation.

This nexus is significant in the evolution of our knowledge infrastructure. It has, for example, catalysed the remarkable success and adoption of large language models (LLMs). Machine learning approaches are rapidly being adopted into our digital research methods, and into the infrastructure itself.

It is interesting to reflect on the various affordances we have identified in the context of scholarly primitives (Unsworth 2000). Some of them might suggest, or at least align with, candidates for new primitives; for example modelling, prototyping, crowdsourcing and training a machine learning model.

We suggest that another new affordance of the digital, and perhaps a scholarly primitive, is play (De Roure and Willcox 2020). Play involves observing, testing and trying; it involves imagination and, crucially, it is safe to fail. Newcomers to programming are often relieved to discover they don’t break the computer if they get something wrong. Play encourages experimentation and immediate learning from failure; digital affordances mean these experiments can proceed at pace.

We can experiment easily, bringing multiple approaches to analyse a dataset and borrowing ideas from near and far: does a method of analysing a corpus of eighteenth-century French literature reveal useful insights when applied to the same genre, but in nineteenth-century Italian? Can a sequence-matching algorithm used in bioinformatics be applied to music analysis? Playing with software also underpins the hackathon as a method, and indeed is celebrated in hacker culture (Parrish 2016) – it is naturally something we can do with software, and naturally something humans do. It catalyses creativity: as Boden says, ‘Creativity is the ability to come up with ideas or artefacts that are new, surprising and valuable’ (Boden 2004). Play as a scholarly primitive is embedded in our scholarly practices and enables failure and innovation at the same time.

We would argue that our enduring infrastructure should also be designed for play, for example through programmatic interfaces. These new affordances empower us to create new infrastructure ourselves, and to share it. This is powerful, though we must be aware of the responsibilities that go with it. When we create infrastructure we are party to it not being agnostic, and with disaffordances in mind too, we should be transparent about the costs: both machine learning and crowdsourcing can be expensive, in energy and the time of volunteers. These are important aspects of Responsible Innovation (Jirotka et al. 2017), in the context of software and of research infrastructure.

Software, innovation and failure

Innovation in software tools and their application occurs right across the disciplinary landscape. This involves experiments in designing and applying digital methods, and these are possible due to the ease of creating and adapting software. Key to this process of experimentation is failure. This is not just tolerated but absolutely fundamental to the process, and the flexibility with which we can conduct these digital experiments means this can be low risk.

Does this apply equally to the sciences and humanities? To a great extent, the computational aspects are applicable, and practices can be shared. This is illustrated well by the practice of open source software development, and code sharing on GitHub. We might view GitHub as a social edition of software, which we are all able to use and to curate, with a comprehensive provenance.

Relatedly, the notion of reproducibility is promoted in the digital research world and is directly relevant to digital humanities and digital scholarship. In this context, reproducibility (which we might think of as ‘computational reproducibility’) means making data and code available so that the computations can be executed again with equivalent results. This is hugely valuable in sharing practice, and it enables outputs to be tested, compared and interpreted, and for new experiments to be conducted – supporting innovation.

However there is a stronger notion, and extensive history, of reproducible research in the sciences, and it is very much about failure. To test a proposed addition to the common knowledge, the scientific method requires independent replication of an experiment – by different researchers, in a different laboratory, typically implementing the same methodology. Replication is essential to give confidence in scientific results, hence repetition and failure are fundamental parts of the process.

This approach to independent replication is perhaps where the established cultures most differ. Science necessarily has a culture of documenting failure and of ‘repeating’, as part of an accepted and principled method. There are many comparisons to be drawn, but traditional humanities scholarship does not typically report failure in this way. We can imagine a scientific reproducibility approach to archival research: do two people get the same result from the same archive, do we build confidence in a result by going into more than one? However, we do see some practice of reproducibility, for example in archaeology where replication studies have educational benefits too (Marwick et al. 2020).

Perhaps, then, what we are proposing with the digital approaches is to support this process of interpretation and model building – augmenting our mental models with digital ones, which give us new means of observing, testing and sharing.

But for the digital part, we can innovate and play, fail and improve, and we can document this through open software practices. By sharing software, computational reproducibility demonstrates a less independent replication than science – it might be compared to sharing a lab or at least apparatus. But this is useful in an ecosystem which is growing in communities of practice and rapidly evolving – where labs are scarce and shared ones welcome.

Permission to be creative: permission to fail

At some point in the future we might lose the ‘digital’ prefix and digital scholarship will just be called scholarship, with the digital methods assumed where they are relevant. But the only way to get to that point is first to develop the digital methods, skills, capacity and career paths. We have shown that innovation in methods involves failing and learning from that failure. What is essential is to give people the power of play, permission to be creative and permission to fail, so that this innovation can occur.

Notes

1. 1 See https://www.software.ac.uk/ (accessed 25 November 2024).

2. 2 See https://documentation.huma-num.fr/humanum-en/ (accessed 25 November 2024).

3. 3 See https://www.ukri.org/news/ukri-continues-investing-in-improving-research-software-practices/ (accessed 25 November 2024).

References

• Berry, David M. and Mark C. Marino. ‘Reading ELIZA: Critical Code Studies in Action’, Electronic Book Review (3 November 2024). Accessed 25 November 2024. https://electronicbookreview.com/essay/reading-eliza-critical-code-studies-in-action/.
• Boden, Margaret A. The Creative Mind: Myths and Mechanisms. 2nd edn. Routledge, 2004. https://doi.org/10.4324/9780203508527.
• Cohen, Jeremy, Daniel S. Katz, Michelle Barker, Neil Chue Hong, Robert Haines and Caroline Jay. ‘The Four Pillars of Research Software Engineering’, IEEE Software 38, no. 1 (2021): 97–105. https://doi.org/10.1109/MS.2020.2973362.
• De Roure, David and Pip Willcox. ‘Scholarly Social Machines: A Web Science Perspective on Our Knowledge Infrastructure’. In Proceedings of the 12th ACM Conference on Web Science (WebSci ’20), 49–58. New York: Association for Computing Machinery, 2020. https://doi.org/10.1145/3394231.3397915.
• De Roure, David and Pip Willcox. ‘Numbers into Notes: Digital Prototyping as Close Reading of Ada Lovelace’s “Note A”’. Paper presented at ADHO 2017, McGill University, Montreal, 2017. Accessed 25 November 2024. https://dh2017.adho.org/abstracts/540/540.pdf.
• Jirotka, Marina, Bastian Grimpe, Bernd Stahl, Gillian Eden and Mark Hartswood. ‘Responsible Research and Innovation in the Digital Age’, Communications of the ACM 60, no. 5 (2017): 62–8. https://doi.org/10.1145/3064940.
• Ma, Bofan, Ellen Sargen, David De Roure and Emily Howard. ‘Learning to Learn: A Reflexive Case Study of PRiSM SampleRNN’. Paper presented at AIMC 2024, Oxford, 9–11 September 2024. Accessed 25 November 2024. https://aimc2024.pubpub.org/pub/fnpykfdv.
• Marwick, Ben, Li-Ying Wang, Ryan Robinson and Hope Loiselle. ‘How to Use Replication Assignments for Teaching Integrity in Empirical Archaeology’, Advances in Archaeological Practice 8, no. 1 (2020): 78–86. https://doi.org/10.1017/aap.2019.38.
• Michie, Donald. ‘Advances in Programming and Non-Numerical Analysis’, Nature 200 (1963): 641–2. https://doi.org/10.1038/200641a0.
• Naughton, John. ‘Lecture: Getting from Here to There’, MedieKultur: Journal of Media and Communication Research 30, no. 57 (2014). https://doi.org/10.7146/mediekultur.v30i57.18609.
• Parrish, Allison. ‘Programming Is Forgetting: Toward a New Hacker Ethic’. Open Transcripts, 2016. Accessed 25 November 2024. opentranscripts.org/transcript/programming-forgetting-new-hacker-ethic/.
• Taylor, Rebecca, Johanna Walker, Simon Hettrick, Philippa Broadbent and David De Roure. ‘Shaping Data and Software Policy in the Arts and Humanities Research Community: A Study for the AHRC’. Zenodo, 2022. https://doi.org/10.5281/ZENODO.10518740.
• Unsworth, John. ‘Scholarly Primitives 20 Years Later’. Video. Version 1.0.0. DARIAH-Campus, 2020. Accessed 25 November 2024. https://campus.dariah.eu/resource/posts/scholarly-primitives-20-years-later.
• Unsworth, John. ‘Scholarly Primitives: What Methods Do Humanities Researchers Have in Common, and How Might Our Tools Reflect This’. Paper presented at the Symposium on Humanities Computing: Formal Methods, Experimental Practice, King’s College, London, May 2000. Accessed 25 November 2024. https://people.brandeis.edu/~unsworth/Kings.5-00/primitives.html.