6.4.1 Standardisation

Standardisation is the accepted approach in the literature for facilitating data management and use. Better and more rigorous standardisation of data improves the efficiency (Ibanez et al., Reference Ibanez, Schroeder and Hanwell2014; Santos et al., Reference Santos2016), reliability and quality of research (Herres-Pawlis et al., Reference Herres-Pawlis2022; Pétavy et al., Reference Pétavy2019) and in general of data-based knowledge production (Gal and Rubinfeld, Reference Gal and Rubinfeld2019), it helps to make data more manageable (Palaiologk et al., Reference Palaiologk, Economides, Tjalsma and Sesink2012; Pétavy et al., Reference Pétavy2019), reusable (Faniel and Jacobsen, Reference Faniel and Jacobsen2010; Herres-Pawlis et al., Reference Herres-Pawlis2022), portable and interoperable (Blobel et al., Reference Blobel, Ruotsalainen, Giacomini, Blobel, Yang and Giacomini2022; Gal and Rubinfeld, Reference Gal and Rubinfeld2019; de Mello et al., Reference de Mello2022; Ribes and Lee, Reference Ribes and Lee2010), and more easily integrated with other datasets (Ibanez et al., Reference Ibanez, Schroeder and Hanwell2014). Standardisation improves the reproducibility of research (Ibanez et al., Reference Ibanez, Schroeder and Hanwell2014), facilitates data discovery (Contaxis et al., 2022) and meta-analysis, that is, synthesising results from multiple studies (Laird et al., Reference Laird2011). Data is also easier and more cost-efficient to preserve if it is well-standardised (Palaiologk et al., Reference Palaiologk, Economides, Tjalsma and Sesink2012). Finally, an often recited benefit is that standardised data is helpful in interdisciplinary collaborations where the different stakeholders lack common knowledge of each others’ data practices (Ribes and Lee, Reference Ribes and Lee2010).

However, while standardisation has many evident benefits, it also has several drawbacks. Archaeological information management literature has for a long time debated the relative benefits of standardisation and allowing researchers to document their data as they find it relevant in their specific contexts of work (Huvila, Reference Huvila, Zhou, Romanowska, Wu, Xu and Verhagen2012a). A major drawback of standardisation is that standards are always approximations and never entirely in line with local practices or understanding of the processes or knowledge that is being standardised (Ellingsen, Reference Ellingsen2004). Standards are not always easily scalable for larger or smaller use cases than was originally envisioned (Hanke et al., Reference Hanke, Pestilli, Wagner, Markiewicz, Poline and Halchenko2021). Also important to note here, as we indirectly discuss data reuse, is that even if retroactive standardisation and data wrangling is possible, standardisation works best when done upfront. Retroactive standardisation tends to be problematic not only because there are seldom resources for comprehensive retroactive working through large datasets but also because of the imminent risks of unintentional data loss, misinterpretations and misrepresentation of original data (Mandell, Reference Mandell2022).

Flexibility is often a practical necessity in highly context specific situations where standards are seldom comprehensive and nuanced enough to accommodate local variation, divergent epistemic perspectives and observations that are difficult to classify into predetermined categories. A study of archaeological heritage administrators’ information work in Sweden provides evidence of how maintaining a fine balance between standardisation of both work processes and documentation, and what is described in the study as ambiguity, is crucial for the success of work (Huvila, Reference Huvila2021). Besides being cumbersome and costly, over-standardisation can become directly counterproductive and an obstacle to both data creation and use. This is an important perspective to hold on to in discussing something as contextual as paradata and one of the reasons for framing paradata management on a spectrum between standardisation and messiness. Keeping this in mind, the three strategies discussed in this section (i.e. intensifying, extending and adjusting granularity in standardisation from Figure 6.1) should be understood as strategies to complement and be complemented by strategies geared towards embracing messiness.

Intensifying standardisation work (Figure 6.1) is suggested in light of the fact that the standardisation of what we term in this chapter core paradata is still very much work in progress. Returning to survey studies and 3D visualisations (previously mentioned as a context where an emergence of core paradata as a documentation category is identifiable), we can say something about the state of paradata standardisation. Survey research has developed a robust corpus of metadata standards but standards for paradata are close to non-existent (Schenk and Reuß, Reference Schenk, Reuß, Huvila, Andersson and Sköld2024). A review of heritage visualisation standards shows that while many standards provide means to represent practice and process information, they have varying emphases and cover the full spectrum of practice and process information only partially. One of the most highly cited documents with direct relevance to paradata in that particular field are charters (including the London Charter, see Denard, Reference Denard2013; and Seville Principles, or the International Guidelines for Virtual Archaeology, International Forum of Virtual Archaeology, 2011; Bendicho, Reference Bendicho2013). The charters provide generic recommendations to document paradata rather than detailed advice on how to do it in practice. Existing metadata schemes and standards cover core paradata, but typically do not use the term paradata (exceptions e.g. CARARE metadata schema, see D’Andrea and Fernie, Reference D’Andrea and Fernie2013). When they cover such paradata-like information, it is to varying degrees and often with a focus on historical processes and data curation.

A survey of major data documentation schemes conducted as a part of the CAPTURE project suggests that the situation is similar in many generic standards. At the same time, as Börjesson et al. (Reference Börjesson, Sköld and Huvila2020) suggest, even if there are few or no dedicated paradata standards, such generic conceptual models as the International Council of Museums (ICOM) International Committee for Documentation (CIDOC) Conceptual Reference Model (CIDOC CRM) provide means to represent practice and process information in detail to an extent that dedicated standards are perhaps not always necessary. Combining complete schemas and individual elements can help to achieve the required coverage of details. First and foremost, there is an urgent need to develop new and to adopt existing data documentation standards to cover paradata and devise guidelines for core paradata across data-producing and data-using communities. All of this work does not need to be exhaustive. For example, many fields have informal standards and vocabularies to describe phenomena interesting to them. Many of them have never been written down and the terms have not been explicitly described. As Huvila et al. (Reference Huvila, Börjesson and Sköld2022) remark, while such terms might be generally comprehensible, a systematic documentation of terminology and an effort to invite data producers to commit themselves to a better standardised use of concepts could provide significant benefits with relatively little overhead.

Expanding on standardisation work (Figure 6.1), not just intensifying the current work in progress, is also sorely needed. We particularly find it advisable to expand standardisation work on two fronts: generally to temporarily cover a larger part of the continuum (or life-cycle) of data; and, more specifically, to formalise the documentation of process information that corresponds to established paradata needs.

Along with the temporal extension, we are suggesting standardising critical elements of core paradata already as a part of routine research documentation. This would require, in practice, more structured advice in the methods literature (cf. Huvila et al., Reference Huvila, Andersson and Sköld2022b; Huvila and Sköld, Reference Huvila and Sköld2023) about how to document key aspects of specific types of data creation and processing activities. In addition, there is a need for greater availability of hands-on advice on what specific paradata should be added to datasets that are produced for sharing and reuse at the planning stage of data collection or generation. This should cover not only what type of information should be added but also more specifically what it is expected to communicate.

As with all data documentation requirements, it should be adjusted to the plausible extent and types of uses relevant to particular datasets. For a lot of data, minimal paradata might be enough whereas datasets with high expectations of meticulous process documentation and high standardised reuse potential should be documented in more detail. With small-scale, highly contextual qualitative data collection with low likelihood of direct reuse of the generated data, a sufficient level of standardised documentation could be a brief narrative of data collection procedures and their underpinning rationale, especially if reusing the data can be expected to require a lot of interpretative work from secondary users. This type of data set and reuse scenario would, however, benefit from such strategies as presented further below, that would prevent the loss of potential paradata that could facilitate that interpretative reuse work. In contrast, there are datasets that require more formal and technical paradata to a much higher degree. This is true for cases such as longitudinal and cross-comparable survey datasets that are produced to be used broadly in both academic and societal knowledge production.

The parallel approach suggested here is equally necessary to starting the generation of core paradata earlier in the data creation process, that is, to formalise the documentation of practice and process information that corresponds to established and common paradata needs. In simple terms, this could mean verifying that the major categories of paradata needs (including those discussed previously in this chapter) are covered to a satisfactory degree in the existing and anticipated core paradata. While doing this it is crucial to consult data producing and using communities to establish what particular information corresponds with categories like methods paradata, scope paradata, provenance paradata and knowledge representation paradata in specific contexts. In such consultations, it is important to both include individuals with a relatively short working experience in the particular field, and those who are more reliable on formal explicit information. Huvila’s study of archaeological heritage administrators suggested the experienced administrators could often compensate for missing explicit information due to their experience of working with the data on a daily basis. In contrast, other stakeholders could not be expected to be able to do that but must rely on the formal standardised documentation (Huvila, Reference Huvila2021). The categorisation presented in this chapter can, in consulting relevant communities, provide a starting point for discussion and a checklist for assessing whether conceivable types of needs have been covered in the documentation requirements.

Adjusting the granularity of documentation (Figure 6.1) is the last of the strategies that are conceptualised as standardisation work in this framework for managing paradata. However, as we shall see, it is placed in the middle of the spectrum for a reason. While determining the necessary level of granularity is certainly difficult, we suggest that it is possible to use the presence and expressed needs of auxiliary documentation as a working heuristic to assess the need of more or less core paradata. If a particular data-creating community is generating a lot of auxiliary documentation that qualifies as paradata, it is probably worth investigating whether some of it might be either replaced by or supplemented with core paradata. In some cases it might be possible if the additional documentation is kept due to the absence of adequate formal documentation standards, making it a question of standardisation work.

In other cases, as in ethnographic research, it is not feasible to replace a notebook with anything that qualifies as core paradata, rather the notebook might be kept or referenced as an identified source of potential paradata (more on this in Section 6.4.2). Correspondingly, repeated demand in data-using communities for either existing or missing auxiliary documentation might signal the need for additional core paradata. This could apply, for example, to the inconsistency identifying data creators or specific measurement instruments in the data documentation. The earlier discussed case of emails and personal information exchange being consulted for augmenting report documentation provides another example of a possible need for augmenting the scope of core paradata. It is seldom feasible to ask for a full email correspondence between researchers to be submitted alongside a dataset to a repository even if it would contain useful information for future reusers of the data. Nor is it reasonable to ask for testimonies or contact information for every single person involved in a research project to be included in report documentation.

However, leaning on our increasing knowledge of key actors and their correspondence being potential sources of paradata, the structures for documenting authors and contributors could be adjusted to increase the likelihood that a reuser can find the right person to talk to in order to better understand, for example, decision-making and contextual premises of making a specific dataset. Here, there are also important opportunities for synergies with the strategy of facilitating discovery and wayfinding, discussed Section 6.4.2 on embracing messiness.

6.4.2 Embracing Messiness

For potential paradata, the usefulness and, indeed, presence of anything deemed paradata will be particularly situational and the only course of action may be to embrace and cater to some level of messiness. It is apparent that cleaning a dataset too much obscures the research process behind it. Literal cleaning is unlikely to happen at a data repository due to the lack of time. Even researchers engage in cleaning proper only selectively. Instead, both data managers and producers alike make numerous decisions of requiring or depositing (only) structured data, leaving out parts of data, datasets or auxiliary information, converting and consolidating that all contribute to the resulting data being a little cleaner than before. It is a result of forcing systematisation of measurements, observations and interpretations, which might not be conclusive enough to be systematised to such an extent. Besides obscuring underlying processes, cleaning also challenges both reliability assessments and several reuse scenarios as discussed above. Of course, it is not possible to keep everything. Rather, as with documentation in general, the question is how to document (or keep) enough, and identify what is most likely to be useful. It is important to try to find a position comparable to what York (Reference York2022) terms ‘reuse equilibrium’ where there is sufficient information but not too much. Drawing on findings on where potential paradata is more or less likely to occur in a dataset in different contexts can facilitate the choice of what to leave messy, what to clean or not to keep. In this part of the chapter we explore how the resulting messiness can be managed and navigated. We propose starting by looking to traditional archiving practices for possible avenues for managing potential paradata. Guided by some of the fundamental principles of Western archiving practices we propose ways of embracing messiness that include a strategy to keep messes messy (i.e. preventing loss of potential paradata) and a strategy to navigate the messiness (i.e. levering linking for paradata discovery).

Preventing loss of potential paradata (Figure 6.1) is rather vague and not particularly actionable without further explicating by what means loss should be prevented. Unsurprisingly, the realisation that it is near impossible to predict what someone in the future will need to know in order to understand the context of some piece of information is far from new – nor is it born from paying attention to paradata. To some extent, a solution to this was proposed within Western archival practice more than a century ago. Having previously applied subject-based classification akin to classification in libraries, archivists in post-revolutionary France developed another principle for arranging and describing records that would be more efficient than the forbiddingly labour-intensive reorganisation of records.

The resulting system, which would be adopted gradually in the Western archival practice throughout the first part of the 1900s, is commonly referred to as respect des fonds. The general idea is that records should be maintained as they were originally accumulated, in their original fonds. A ‘fonds’ is all the records of, for example, an administrative authority, or a family, or a person (Trace, Reference Trace, Duranti and Franks2015, p. 21–24). Two important principles inform how records should be organised within fonds, that is, inform archival arrangement and description; these are the principle of provenance and the principle of original order. As discussed in Chapter 2, the principle of provenance puts emphasis on the importance of relationships between records and the context, functions and agents (e.g. organisations or individuals) involved in their creation, accumulation and use. The principle of original order promotes maintaining the ordering structure established by the records’ creator in order to preserve important contextual information.

Inherent in the principle of original order is a view of the archival records as evidence of the activity of a particular organisation. The idea of respect des fonds, with its underlying principles, becomes rather more complicated when applied in practice, since the application of any standardised model of these principles, like Oliver Wendell Holmes’s (1988) influential nested hierarchical model, will impose structure on the original order. However, the foundational idea of respect des fonds offers valuable insights that can be applied to the management of paradata, particularly for endeavours to preserve potential paradata.

The implication of respect des fonds and its parallel principles of original order and provenance for managing potential paradata is to keep the fonds of such information intact to an extent that is possible. From a repository perspective, potential paradata is essentially unmanageable when it falls outside the formal structure of how a data repository deals with datasets and metainformation, describes and preserves them, and works towards making and keep them formally FAIR or findable, accessible, interoperable and reusable (Wilkinson et al., Reference Wilkinson2016). Focusing efforts on keeping such information technically readable, and documenting it only superficially to maintain a rudimentary degree of findability is from a repository perspective obviously a question of managing resources. If keeping the information is useful enough, it is possible to avoid costs of turning it into core paradata.

However, as the previous discussion in this volume especially in Chapter 3 shows, it also has qualitative benefits. Turning potential paradata into core paradata is not only costly but it carries also a risk of losing information potentially available in the organic, unrefined messiness of such resources. For example, with methods paradata, looking into the terminology used to describe methods as a whole is informative not only of specific methods applied but also of how research practice and process is conceptualised as a whole. Moreover, as the work of Börjesson et al. (Reference Börjesson, Sköld, Friberg, Löwenborg, Pálsson and Huvila2022) shows, comparing earlier and later versions of interpretations in datasets can help to trace interpretative processes otherwise undocumented in the available research material.

Having established that some messiness may be beneficial for the preservation of potential paradata, the next question becomes how we might manage such a mess in a relatively systematic way. Again, we turn to the basic principles of archiving and, more specifically, how the idea of archival finding aids may be adapted to be useful tools for paradata management.

Facilitating paradata discovery (Figure 6.1) is a strategy that takes as its starting point archival finding aids. McNeil offers a useful definition of an archival finding aid as ‘any tool that aims to provide users with intellectual and/or physical access to the holdings of archival institutions’ (MacNeil, Reference MacNeil2012, 486). This definition can easily be reframed in a data and paradata context, as a tool that aims to provide users intellectual access to the practices and processes underpinning a particular assemblage of data.

Taking directions from traditional archival practice in this way we can introduce some structure to how we can navigate the messes with paradata potential. There are obviously different means to implement a functioning paradata finding aid. Similarly to paradata itself, finding aids are also contextual. However, in contrast to paradata itself, considering the frame within which they are used, finding aids can be expected to be contextual first and foremost in relation to future data use rather than data creation. In parallel, rather than focusing on what such a finding aid should be, it is more fruitful to plan for what it should do (cf. MacNeil, Reference MacNeil2012) for data creators, managers and users. It is important to consider earlier evidence of the advantages and affordances of traditional paper-based and digital finding aids when considering their potential adoption to paradata. A finding aid can facilitate paradata discovery through providing links between information resources, keeping traces of previous use of the data and the finding aid itself.

Perhaps the most significant function of a paradata finding aid would be to provide directions. Because all types of potential paradata require a lot of interpretative effort to use, it is superfluous to attempt to describe it beyond what is necessary to use the tool itself. Explicating the aboutness of datasets is crucial for understanding their processual context similarly to providing directions where to find information that helps to interpret it, including vocabularies and descriptions of colloquialisms. A finding aid should also provide an overview of the principles of how data was organised and made, and what standards were used and how un-FAIR as in unfindable, unaccessible, uninteroperable and unreusable the data is (Börjesson et al., Reference Börjesson, Huvila and Sköld2022). In parallel, a good finding aid would provide directions across contextual and disciplinary boundaries. In many cases relevant data and paradata is not only disconnected in different logical units within one repository but literally scattered around the world. An illustrative example of this is historical and archaeological data that in many cases has been scattered to museums and research institutions located on different continents (e.g., Börjesson et al., Reference Börjesson, Huvila and Sköld2022; Sobotkova Reference Sobotkova2018; Stilborg Reference Stilborg2021). Remedying or at least bringing partial alleviation to the greatest obstacle in mediating practice and process information, namely its contextuality, is crucial for any tool proposed for the task. While a universalist ideal a lieux Otlet’s Mundaneum or Bush’s Memex would be a complete meta-level inter-institutional directory to (all) paradata, much more modest and context-specific directories could if not solve the paradata problem as a whole, at least alleviate it considerably.

However, rather than going entirely contextual, there would also be possibilities to develop standards for formats and contents of finding aids for potential paradata. Solving the problem of diverging expectations of data managers, creators and data reusers and what is feasible to achieve for paradata finding aids requires reaching a compromise between what (all) would be regarded as desirable. Such standardisation would be useful for potential reusers seeking potential paradata specific to their needs, but would also be helpful for data creators in providing rudimentary structure to data documentation efforts. Moreover, as Battley (Reference Battley2013) reminds us, archival findings aids are also records, or with paradata, another layer of paradata describing data management practices and processes (cf. Gant and Reilly, Reference Gant and Reilly2017). Generally, the standardisation of finding aids for paradata would also support paradata literacy, which is the topic we will now turn to in the final part of this chapter.

6.4.3 Paradata Literacy

There can be no doubt that solving the conundrum of generating or identifying paradata that is useful for its users, feasible to generate and keep for data creators, and manageable from a repository perspective is not merely an issue of selecting suitable technologies and particular types of resources to solve the problem. Paradata as a type of information, about processes and practices, is a moving target. The most comprehensive method for managing paradata may be simply to acknowledge that there is a need for process information not currently taken into account in documentation strategies and that information plays a much more critical role in the chain of knowledge-making and use than a mere footnote or an auxiliary attachment to data proper.

In contrast to finding aids that provide a map to existing paradata and guidance how to find and navigate it, we suggest looking to paradata literacies as a crucial step in making paradata (proper and potential) actionable in context and especially in working out the conundrums of identifying and making diverse forms of potential paradata. This discussion is both complementary to, and more overarching than, the discussion of paradata usefulness thresholds in Chapter 3. Finding aids and the notion of paradata literacy approach the paradata challenges from different angles. Finding aids can help make paradata and the practices and processes it documents better findable and accessible. Paradata literacies are needed to make data and data practices and processes work both on their own and in relation to each other. In practice, both have a much broader scope and in an ideal situation they would complement each other in a pursuit of increasingly relevant, resonant, ethical and sustainable paradata practices, practices that support corresponding, reflective and epistemically inclusive and sensitive data practices.

The argument for increased mutual paradata literacy is based on the insight that the discrepancy between available and desirable paradata is not simply an information problem. It is also, at least to an equal extent, a problem of disparity in paradata practices and competences, as observed in Chapter 3 where the issue is termed the ‘epistemic usefulness threshold’. Researchers have been criticised in many studies for a lack of adequate competence in data management (e.g., Pálsdóttir, Reference Pálsdóttir2021). The same critique could be undoubtedly extended to paradata competences, although as our findings from a survey and interviews on researchers’ paradata practices suggest, data creators, managers and users alike show a high degree of often tacit competence in sharing and acquiring practice and process knowledge. The trouble is that these competences, or literacies, do not necessarily converge.

Contemporary data repositories have structures in place to cater to many of the researchers’ paradata needs, and data managers have skills and capacity to facilitate paradata practices but researchers are not always knowledgeable of what they are, how the existing structures and tools work, and how the available resources could be utilised for generating and preserving better paradata. Data managers are similarly struggling with securing a sufficient understanding of how data is created and used to be able to help researchers and other data users and producers. Therefore we want to underline the importance of improving mutual paradata literacy (Figure 6.1) instead of focusing on individual stakeholder groups.

Paradata literacy is a natural extension to what has been discussed in terms of data literacy or literacies (Huvila et al. Reference Huvila, Kaiser, Sköld and Andersson2024; Reference Huvila, Olsson, Sköld, Kaiser and Andersson2025). It can unfold as something comparable to the competence and associated abilities that have been termed metadata literacy (Mitchell, Reference Mitchell2013) and metadata literacy skills (Çakmak and Kurbanoğlu, Reference Çakmak, Kurbanoğlu, Kurbanoglu, Boustany, Špiranec, Grassian, Mizrachi and Roy2015). It can also be seen as a complement to data literacy, essentially framing paradata literacy as a key facet of established data literacies. Paradata literacy, in the sense relevant here, would cover the competences of identifying, generating and keeping relevant and adequate core paradata and understanding what it does for data and data use.

The major implication of approaching the conundrum of managing paradata as a literacy question is that of acknowledging that making paradata work requires competence and effort and qualifies as a specific type of undertaking of its own. Data managers, creators and users all have specific roles and need to develop systematic competence to contribute to enacting paradata in practice. Acquiring and developing data management skills identified in the literature helps to develop that competence (e.g., Koltay, Reference Koltay2015; Schneider, Reference Schneider, Kurbanoğlu, Grassian, Mizrachi, Catts and Špiranec2013; Vilar and Zabukovec, Reference Vilar and Zabukovec2019). However, considering the often indefinite boundaries of what paradata is and where to find it, paradata literacy extends beyond being a simple skill set. Paradata literacy is first and foremost a matter of understanding the complexities and challenges of working with paradata where theoretical, ethical and interpretative facets of practice cannot be detached from technical skills and practical doing (cf. the ‘technical usefulness threshold’ in Chapter 3 and Kansa and Kansa, 2020). It links to the broader field of data literacies (Verdi and Deuff, Reference Verdi and Deuff2021) sharing many of the identified challenges but also opportunities and ways forward (cf. Koltay, 2021).

Like data literacies, paradata literacies are also never value-free. In their complexity and embeddedness in data practice, paradata literacies unfold rather as assemblages of ongoing processes and practices that need to evolve in time and across diverse contexts of data creation, management and use. Paradata literacies are also necessarily reciprocal in the sense that the making of paradata needs to be aligned with its management and use, and vice versa. being a competent paradata creator, manager or user, requires adeptness in the whole lifespan of paradata from its inception to management, use, reshaping and beyond.

Similarly to other literacies, paradata literacy is also about finding a balance between empowering users and providing them relevant services (cf. Huvila, Reference Huvila2012b). The analysis of Weber et al. (Reference Weber2023) points to a similar gap between archival users and archivists that is evident also in the broader sphere of data management. It is likely that much of future paradata use will not be mediated by data managers but rather by data creators and users themselves who are depositing and accessing datasets directly in online repositories.

However, even if this would be the case, there is still need for professional support, probably to a greater extent than is acknowledged in the contemporary management culture, which often downplays the significance and both qualitative and economic advantages of qualified professional service work. We are still to see a well-grounded division of labour between what users should be able to achieve independently and what types of services are necessary and appropriate to offer and when. Repositories and data managers occupy a position with a broad vista across diverse context-specific data practices. In doing so, they hold a unique position and responsibility to not only act as service-providers but to actively raise questions and act in dialogue with data creators and users to find virtuous paths forward.