2.3.1 Paradata in Survey Research

There are indications that the earliest academic use of the concept of paradata took place in survey research, although the roots of the concept and particularly the data frame of mind it encapsulates are unclear – see for instance Huvila (Reference Huvila2012) referencing Adkins and Adkins (Reference Adkins and Adkins1989) as an early example of paradata-related discussions of how to record past data processes in analogue settings. While Lyberg (Reference Lyberg2009) has traced the origins of the term paradata even further back to the 1920s, studies that seek to find a core paradata definition commonly reference two papers by Couper from the turn of the millennium (Couper, Reference Couper1998 and Reference Couper2000, the former as cited in Kreuter and Casas-Cordero, Reference Kreuter and Casas-Cordero2010). Working with the issue of how to improve the usability of computer-assisted survey methods, Couper defines paradata as ‘auxiliary data describing the [survey] process’ (Couper, Reference Couper2000, p. 393). Paradata here is described as data that can provide insight into the work of enacting surveys, manifesting for instance in (para)data telling about the survey procedure (the no. of interviewer calls per response, response rates, keystroke logs showing how the interviewer managed the survey system) and the survey results (e.g., interview length per average). Couper (Reference Couper2000) also suggests that useful paradata can trace data from web surveys that would allow evaluation of different parameters that together show how the respondents interacted with the survey website or tool.

Couper’s conceptualisation of paradata has seen widespread use both in survey research (e.g., Nicolaas, Reference Nicolaas2011; Olson, Reference Olson2013) and in other domains (e.g., Huvila et al., Reference Huvila, Greenberg, Sköld, Thomer, Trace and Zhao2021a; Reilly et al., Reference Reilly, Callery, Dawson and Gant2021), and it has three key characteristics that are recurrent in a large portion of academic paradata use cases. One such characteristic is that paradata has an ‘auxiliary’ or supporting role in relation to another dataset that is the principal outcome of the practices or processes described by the paradata, like the survey results in Couper (Reference Couper2000). Another characteristic is that paradata is quite intimately tied to the wide-ranging documentation affordances of technology, where large amounts of data of different granularities can be recorded that describes system interactions almost in parallel to the interactions taking place. A third characteristic is the multimodal temporality of paradata. Paradata is directed towards the past in the sense that it informs about past events and processes. Paradata is, however, also tied to the present and future in its strong utilitarian connotations – paradata is sought because it informs about past events and processes in a way that facilitates present or future actions, evaluations or interpretations in connection to the main product or principal outcome that the paradata describes.

Subsequent survey methodology research has continued to develop the concept of paradata and associated methodologies in several directions, emphasising different facets of the concept and its applications and uses (Nicolaas, Reference Nicolaas2011). One significant development is that paradata began to shift into something not solely a ‘by-product’ (Kreuter and Casas-Cordero, Reference Kreuter and Casas-Cordero2010, p. 2) of the survey process recorded by the survey software, but also as a type of data collected using multiple means in service of a broader range of goals and intentions. Beyond offering information about the processes underpinning collected data, paradata was seen to be relevant for understanding missing data and data not collected, like survey non-responses, and gauging respondent interest and involvement in the survey instruments (Couper and Kreuter, Reference Couper and Kreuter2013; Olson, Reference Olson2013). This more comprehensive definition of paradata further included different types of qualitative data generated both by the data creators themselves and the technologies involved in the research work. Paradata became inclusive of observations by the interviewers about, depending on the present research objective, not only the survey process itself and how the respondent approached answering questions, but also pertaining to the appearance of the respondents and impressions and observations stemming from the broader circumstances of the interview (e.g., state of the respondent’s domicile or neighbourhood) (Durrant et al., Reference Durrant, D’Arrigo and Steele2011; Edwards et al., Reference Edwards, Goodwin, O’Connor and Phoenix2017). This type of paradata, manifested commonly in field notes and audio recordings, could include estimations of the respondents’ intellectual process, like the extent to which they were certain about their answers and what, and if so how, documentation was consulted by the respondent when providing responses to queries (Nicolaas, Reference Nicolaas2011).

The concept of paradata was also refined in how it connects to paradata use, driven in part by technological innovation which provided additional and more intricate approaches to creating paradata. For example, eye-tracking methodologies were implemented to record paradata of respondent-computer survey interactions (West, Reference West2011). As available trace paradata in computer-driven survey systems began to increase qualitatively and diversify quantitatively, procedures around how to interpret the resulting data developed. Stieger and Reips (Reference Stieger and Reips2010) and Sharma (Reference Sharma2019) point to the many challenges involved in knowing how log files from systems with different functions in the infrastructures underpinning digital survey systems can be used to capture respondent’s often dynamic and idiosyncratic modes of survey-response behaviours. Further expanding the boundaries of the concept of paradata and its use in the survey settings, there have also been calls to delve into existing datasets in search of paradata that can usefully inform researchers about data-related processes. This paradata, what is termed ‘retrospective paradata’ in Chapter 7, may originally have been recorded for other purposes but can in the current contexts of use provide valuable insights into past data procedures of data creation, curation, and use (Edwards et al., Reference Edwards, Goodwin, O’Connor and Phoenix2017; cf. Börjesson et al., Reference Börjesson, Sköld, Friberg, Löwenborg, Palsson and Huvila2022a).

2.3.2 Paradata in Archaeology and Heritage Visualisation Research

The concept of paradata as it emerges in both archaeology and heritage visualisation research is somewhat similar across the two fields, although differences exist. An important point of common reference is The London Charter, a collection of six principles seeking to promote modes of work in computer-driven visualisation efforts that ensures methodological robustness and transparency, and also advance computer-based visualisations as a means to manage and inquire into heritage and archaeological data (The London Charter Organization, 2009b). Principle 4 in the charter outlines how visualisation methods and the resulting visualisations should be documented. Sub-principle 4.6 is titled ‘Documentation of process (paradata)’ and reads as follows:

Writing about the processes and discussions leading up to the establishment of The London Charter, Beacham et al. (Reference Beacham, Denard and Niccolucci2006) characterise the issue of documentation as being among the most difficult challenges addressed in the charter. The difficulties stemmed not principally from attaining a sufficient degree of methodological transparency via documenting the steps and decision-making involved in visualising heritage and archaeological data, but rather to greatly varying methodological expectations rooted in the methodological traditions of the many disciplines and fields of research that engaged in the production and use of computer-based visualisations. Paradata was suggested by Denard (see Denard, Reference Denard, Bentkowska-Kafel, Denard and Baker2016) as a way of denoting descriptions of the mental and physical activities involved in creating the visualisations, complemented by articulations of the intellectual context and implicit assumptions and principles, that might serve to bridge methodological frameworks (Beacham et al., Reference Beacham, Denard and Niccolucci2006). Beacham (Beacham Reference Beacham2011, p. 51 see also Beacham et al., Reference Beacham, Denard and Niccolucci2006) stresses that paradata are descriptions of ‘intellectual capital’ being operationalised in the making of computer-based visualisations, further broadening the scope of paradata to possibly include representations ranging from discrete tasks and strategies of action to epistemic horizons.

The London Charter is widely referenced in archaeology and heritage visualisation research discussing paradata (e.g., Huggett, Reference Huggett, Mills, Pidd and Ward2012; Morgan and Winters, Reference Morgan and Winters2015; Niccolucci et al., Reference Niccolucci, Felicetti, Amico and D’Andrea2013), although it has been noted that paradata in the archaeological context is not well defined (Huggett, Reference Huggett2020), and that the principles of the charter require adaptation to the specific circumstances and objectives of archaeological heritage, as done in the Seville Charter (Lopez-Menchero and Grande, Reference Lopez-Menchero and Grande2011). Other and complementary conceptualisations of paradata have been suggested that, for example, also highlight that paradata should include descriptions of the evidence (literary sources, image resources, archival data) involved in creating, curating and using archaeological data and heritage visualisations beyond activities and decisions (Barratt, Reference Barratt2016; D’Andrea and Fernie, Reference D’Andrea and Fernie2013). Barratt (Reference Barratt2016), referencing Dell’Unto et al. (Reference Dell’Unto, Leander, Dellepiane, Callieri, Ferdani and Lindgren2013) further expands upon the concept of paradata by categorising different types of evidence (testimonial sources, sources based on current measurements and analyses) and intellectual processes (deduction, comparison, hypotheses) that are useful to include when describing past processes of scholarly work. Exploring paradata from a data management perspective, Kansa et al. (Reference Kansa, Atici, Kansa and Meadow2020; cf. InterPARES Trust AI, forthcoming) determine that information about the data authors and their areas of expertise and training are paradata that can be employed to better understand the data or scholarly product created. Kansa et al. (Reference Kansa, Atici, Kansa and Meadow2020) also underline that it is important to also think about paradata that is more closely associated with the data itself to be of wider scope than the data and include, for example, sampling biases and information about missing data (cf. Ullah, Reference Ullah2015) that can help explain why the data appears in the way it does.

Paradata as it is defined and used in archaeology and heritage visualisation research has several similarities to how paradata is approached in survey research, but there are also other conceptual trajectories emerging. The characterisation of paradata as an auxiliary data type vis-à-vis the ‘principal’ data or visualisation scholarly outcomes remains, however, with a slight difference in framing. The general understanding of scholarly work as it is represented in this literature resembles a network of interacting research practices, methodologies and varying kinds of data and epistemic circumstances, where the specific relationships between these network components become important due to how they impact each other and ultimately the nature of the scholarly work as a whole (e.g., Denard, Reference Denard, Bentkowska-Kafel, Denard and Baker2016; Richards-Rissetto and Landau, Reference Richards-Rissetto and Landau2019). Paradata, in reflecting how research is enacted and data is produced in practice, emerges as a data type that encompasses a wide range of key elements and their interplay (see Reilly et al., Reference Reilly, Callery, Dawson and Gant2021), from qualitative paradata describing how understanding and interpretation has happened (D’Andrea and Fernie, Reference D’Andrea and Fernie2013) to operational trace data and data detailing machine settings (Reilly et al., Reference Reilly, Callery, Dawson and Gant2021). While the concept of paradata remains closely associated with technology and technology’s capacity to record representations of how it is used and calibrated (Niccolucci et al., Reference Niccolucci, Beacham, Hermon, Denard, Artusi, Joly, Lucet, Pitzalis and Ribes2010), it is in archaeology and heritage visualisation research to a notably higher degree associated with descriptions of the practices and processes engaged in by the makers of scholarly data and their epistemic attributes. Here, the principal paradata stakeholder group is made up of researchers or other parties seeking to gain insight into data processes by consulting paradata representing the original setting of data creation in a broad-spectrum way, while paradata in survey research commonly reflects the activities of the respondent stakeholder group and how they impact the survey data outputs (Ballin et al., Reference Ballin, Scanu and Vicard2006; Nicolaas, Reference Nicolaas2011).

2.4.1 Metadata

The term ‘metadata’ is said to have originated in the 1960s within the field of computer science (Furner, Reference Furner2020) and has since gained widespread recognition, being utilised in numerous and varied contexts. Metadata is firmly established in the LAM (libraries, archives, museums) sector and associated disciplines (information studies, archival studies, museum studies), where it principally refers to the resources used to describe the attributes of physical and digital collections and holdings so their items can be made searchable, findable, and possible to use and manage and preserve (Mayernik, Reference Mayernik2020; Ronzino et al., Reference Ronzino, Hermon and Niccolucci2012). Gilliland (Reference Gilliland and Baca2008) and Pomerantz (Reference Pomerantz2015) argue, however, that metadata in actuality predates the coining of the term, and that its manifestations have been in existence since the earliest attempts at organising information. The increasing pervasiveness of digital technology in all domains of leisure and labour has been a driver in metadata becoming a colloquial term and a ubiquitous and impactful phenomenon, indispensable in the operation and handling of digital data and digital communications, including social media and other digital platforms and infrastructures (Pomerantz, Reference Pomerantz2015; Zeng and Qin, Reference Zeng and Qin2016).

The literature offers several conceptualisations of metadata that illustrate the purposes that metadata is envisioned to fulfil. Metadata’s fundamental function is to in varying ways enrich the form or forms of information it is appended to by facilitating a range of actions. Some of these actions have to do with knowledge organisation and the work of ordering and describing items, data, books and records according to certain principles and systems that support information search (Mayernik, Reference Mayernik2020) and required administrative and managerial tasks (Kalová, Reference Kalová2020; Tompkins et al., Reference Tompkins, Honick, Polley and Qin2021), including repository storage and long-term preservation (Day, Reference Day2002). Metadata also enables both reuse, that is, use by a party external to an item’s context of creation) and use of the item described by offering insight into its features and attributes (Gilliland, Reference Gilliland and Baca2008) and relationships to related items (Force and Smith, Reference Force and Smith2021). Metadata supports the usability and reusability of forms of information by being a resource that users can employ to validate and contextualise the items, and as evidence in efforts to determine their authenticity, integrity and trustworthiness (Tennis, Reference Tennis2008). Conversely, metadata is also an essential tool in the successful sharing of data, documentation and other forms of information (Chao, Reference Chao2014). Metadata is, however, no silver bullet in facilitating sharing and reuse. It is necessary that the metadata shared is the metadata needed by the stakeholders set to reuse what is shared (Kim, Reference Kim2021), and there might be infrastructural and epistemic challenges that impact their ability to apply the metadata (Hansson and Dahlgren, Reference Hansson and Dahlgren2021).

2.4.2 Provenance and Provenance Data

Like metadata, the concept of provenance is employed in an extensive array of domains and contexts – from computer science to archival science and beyond. Its modes of use and the connotations it carries, including the significance of provenance and the utilities of provenance data, also varies. Provenance has strong connections to the archival field, where it is operationalised in, among other things, the principle of provenance. The principle of provenance is a key tenet of archival theory that describes how archival holdings should be organised by keeping the records from one record creator together and separate from records from other record creators, in this way preserving the meaning and context of the records (Michetti, Reference Michetti and Lemieux2016; Sweeney, Reference Sweeney2008). Provenance also sees use in other fields and disciplines concerned with tracking and documenting data trajectories, and using information about past interactions with forms of information (again, data, records, books, manuscripts) to further present analytical or interpretative efforts (see e.g., Lemieux, Reference Lemieux and Lemieux2016). Provenance is an important term in the field of computer science, in particular in scholarly efforts to better model and describe data workflows and the management of scientific data (Curcin, Reference Curcin2017; Doerr and Theodoridou, Reference Doerr and Theodoridou2011). Additional examples of settings where provenance and provenance data is a concern includes archaeology (Huggett, Reference Huggett, Mills, Pidd and Ward2012), information studies (Fear and Donaldson, Reference Fear and Donaldson2012), museum studies (White, Reference White, McDonald and Michael2017), geology and palaeontology (Reilly et al., Reference Reilly, Callery, Dawson and Gant2021), and art history and cinema studies (Bernardi et al., Reference Bernardi, Usai, Williams and Yumibe2021).

While being used in many settings for a range of purposes, provenance data is broadly conceptualised as a resource principally for making visible how forms of information travel between stakeholders and repositories, and to describe the impacts that these travel pathways have had on the forms of information themselves (Gehani et al., Reference Gehani, Ahmad, Irshad, Zhu and Patel2021; Lemieux, Reference Lemieux and Lemieux2016; MacNeil, Reference MacNeil2008). The literature shows that the information about the ‘histories’ of different forms of information provided by provenance data is understood to be useful for a range of more specific reasons. In settings of use more closely affiliated with the data-centric understandings of the concept, provenance data is seen to facilitate examinations of different kinds, including data audits and validations (Davidson and Freire, Reference Davidson and Freire2008). Curcin (Reference Curcin2017) and Ludäscher (Reference Ludäscher and Lemieux2016) outline how provenance data also can be used to assess results and outcomes by facilitating replicability and reproducibility of data tasks and processes, including both analytic and data management operations. Provenance in the data sciences commonly represent very fine-grained representations of data interactions. Provenance graphs and other provenance accounts can provide detailed insight into the discrete steps involved in data creation, curation and use (Davidson and Freire, Reference Davidson and Freire2008; Doerr and Theodoridou, Reference Doerr and Theodoridou2011).

The use of provenance in disciplines and contexts more aligned with the archival discourse shows many similarities, but also some differences in emphasis and in how the provenance data is manifested. Similarities include using provenance data to drive data exploration (Davidson and Freire, Reference Davidson and Freire2008) and the determination of authenticity and other qualities, including accuracy and relevance (Fear and Donaldson, Reference Fear and Donaldson2012; cf. Davet et al., Reference Davet, Hamidzadeh and Franks2023). Provenance data is also used to better understand the origins of the data or record at hand, inclusive of both actors involved and the broader contexts in which the materials emerged or were created (Buchanan, Reference Buchanan2016; Michetti, Reference Michetti and Lemieux2016), together with impactful epistemic and theoretical circumstances (Huggett, Reference Huggett, Mills, Pidd and Ward2012). However, provenance data in the archival setting less often pertains to the objects or records it describes on the item-level, instead existing predominantly at the collection-level. This makes it difficult, often impossible, to base estimations of authenticity and trustworthiness on detailed documentation of data interactions. Instead, an important function of provenance data is to show chains of custody and ownership that can be used to assess relevance and task-related value (Lemieux, Reference Lemieux and Lemieux2016). It is also often highlighted that provenance data underpins all basic categories of archival work and functionality (arrangement, description, acquisition, retrieval, appraisal, Lemieux, Reference Lemieux and Lemieux2016; Michetti, Reference Michetti and Lemieux2016; Sweeney, Reference Sweeney2008), including preservation (Li and Sugimoto, Reference Li and Sugimoto2014).

2.5.1 Encompassing Paradata, Provenance Data and Metadata

The challenges of encompassing paradata notwithstanding, this chapter has shown that the concept has several core characteristics and also key correspondences to metadata and provenance data. General features of all three concepts are that they – like data and documentation, see Chapter 3 – exist through and in relation to conditions that are both social and material in nature, consisting of, to different extents, field-specific and field-overarching elements of scholarly practice, tools and systems, and epistemic frameworks. Paradata, provenance data and metadata are thus used and produced by both human and non-human actors by automatic or manual means, and they are, to a large extent, system-dependent in the sense that access, management and storage extensively relies on the purposeful functioning of supporting information systems (Force and Smith, Reference Force and Smith2021; Hansson and Dahlgren, Reference Hansson and Dahlgren2021; Kim, Reference Kim2021).

The literature also shows that all three data types can be represented ‘in’ data files and other forms of information, or exist in resources that are distinct from them – like metadata schemas or method descriptions in associated publications (De Oliveira et al., Reference De Oliveira, Silva and Mattoso2015; Pomerantz, Reference Pomerantz2015). Also, the discussed data types emerge in different forms and in different places: They can be highly structured (found in standards, data models) but also unstructured (present in datasets, work logs or notes) and comprise narrative or trace data. In terms of their temporal orientation, the concepts can be prescriptive and determine what paradata, provenance data or metadata should be documented when carrying out a certain task or process. They can also be retrospective and be used to discern and describe data interactions that are in the past or prospective and focused on speculating or stipulating future activities (Chapter 7; Cuevas-Vincenttin et al., Reference Cuevas-Vincenttin2016; Davidson and Freire, Reference Davidson and Freire2008; Ludäscher, Reference Ludäscher and Lemieux2016). Finally, paradata, provenance data and metadata are ‘meta’ concepts (cf. Schenk et al., Reference Schenk, Dividino, Staab, Freire, Missier and Sahoo2009) in that they express the relationships and attributes of forms of information and other data or descriptive resources related to these items.

The relationship between paradata, provenance data and metadata on one hand and the forms of information that they describe on the other, is arguably – at least from a conceptual perspective – both their main designator and possibly their lowest common denominator. This chapter shows that the differences between the concepts are notable. The use cases of paradata, provenance data and metadata reviewed above support some observations about the relationship between the concepts as they manifest in research practice. A general dissimilarity between metadata and paradata and provenance data is that the former term describes the attributes of forms of information, while the latter two terms describe their histories of creation, curation, use and custody. In situations where the concepts are not considered to be distinct, metadata is most commonly attributed the highest order of abstraction. In these instances, provenance data (da Cruz et al., Reference da Cruz, Paulino, de Oliveira, Campos and Mattoso2011; Curcin, Reference Curcin2017; Malik et al., Reference Malik, Nistor and Gehani2010) and paradata (Dahlström and Hansson, Reference Dahlström and Hansson2019; Niccolucci et al., Reference Niccolucci, Felicetti, Amico and D’Andrea2013; Pomerantz, Reference Pomerantz2015) are organised as metadata sub-types.

This is the only observable hierarchical relationship between the concepts. Paradata and provenance data are at times posited as alternatives to each other (Huggett, Reference Huggett, Mills, Pidd and Ward2012), but several authors consider the concepts to be distinct, albeit closely related. The principal difference between paradata and provenance in these cases is that provenance has a stronger emphasis on the technical circumstances of data interactions. Paradata is instead seen to highlight the activities, processes and frameworks involved in interpreting and understanding data (D’Andrea and Fernie, Reference D’Andrea and Fernie2013; Reilly et al., Reference Reilly, Callery, Dawson and Gant2021).

The conceptual relationships most frequently discussed in the reviewed literature are those between metadata and paradata, and metadata and provenance data. Li and Sugimoto (Reference Li and Sugimoto2014) also suggest that metadata creation and description should be considered from the viewpoint of provenance, and that it would be valuable to record and keep. Reilly et al. (Reference Reilly, Callery, Dawson and Gant2021, p. 458) contrasts metadata, described as ‘generally uncontentious static properties of data’, to the more heterogeneous collection of data inputs and byproducts, scholarly activities and interpretative resources captured as paradata.

Studies in survey research show fairly uniform distinctions between paradata and metadata. Metadata is exemplified as providing information about survey-supporting elements, for example, questionnaires, sample designs, settings and configurations, and the web browsers and operating systems used (Barratt, Reference Barratt2016; Olson, Reference Olson2013; Stieger and Reips, Reference Stieger and Reips2010). This is in contrast to paradata, which describes both how the surveys have been interacted with (scrolls, clicks, typing, call records) and the intellectual work involved in enacting the survey study (Barratt, Reference Barratt2016; Stieger and Reips, Reference Stieger and Reips2010). It is interesting to note that the relatively homogenous characterisation of paradata in survey research stands somewhat in contrast to approaches in other fields, where paradata is understood to encompass also the machinery, infrastructures, and tools involved in data creation, curation and use alongside information about their configurations (Huvila, Reference Huvila2022; Kreuter, Reference Kreuter2013).

2.5.2 The Concept of Paradata

In conclusion, the concept of paradata as discussed in the literature is diverse. This chapter has shown that paradata encompasses a range of meanings and definitions. It is also conceptualised as being useful in relation to a various set of purposes and intents. Paradata’s empirical referents – that is, the standards, data models and data or process descriptions that exemplify and instantiate it – are plentiful and assorted. Paradata is connected in close and complex ways to metadata and provenance data, and while some general observations can be made about how the concepts and their modes of use differ and are alike, it is difficult to draw any definitive boundaries between them. This does not mean, however, that there are not any purposeful approaches available that can be used to grasp the concept of paradata and to discern its possible implementations.

One approach is to consider paradata in a more abstract way. From this perspective, it is possible to determine that paradata is a concept that can be used to examine any information or data phenomenon in any work or leisure setting, although it has been most prominently used to examine scholarly settings and scholarly data in survey research, heritage visualisation research and information studies. Paradata is data about the full range of activities, resources and epistemic frameworks involved in creating, managing and using scholarly data. It can emerge in any form or format, and be created intentionally or as a secondary outcome of scholarly data interactions and processes. Paradata is useful for gaining insight into past data trajectories, processes and events. Purposeful paradata can facilitate a wide range of data management and reuse scenarios, and is an asset in examining data when assessing its relevance, reliability, accuracy, scope and content. From a more abstracted perspective, paradata differs from metadata in that it describes data processes, and metadata describes data. Paradata and provenance data are more alike in that both concepts are broadly concerned with the origins and processing histories of data, but provenance data is more strongly situated in the settings of archives, archaeology, art history and computer science which, as we have seen above, impacts what the data looks like, and how it is created and used.

Another approach is to consider paradata in a more tangible way, which also introduces significantly more changeability to the concept. In this sense, paradata emerges as a concept that is differently defined and applied across research settings and disciplines. In survey research, paradata showcases well-regimented expressions and modelings of past data trajectories and interactions that are akin to the computer science workflow-provenance research and standard-driven metadata descriptions. In other domains, paradata is a notably more fluid concept used as an exploratory and interpretative device (see e.g., Börjesson et al., Reference Börjesson, Sköld, Friberg, Löwenborg, Palsson and Huvila2022a; Reilly et al., Reference Reilly, Callery, Dawson and Gant2021). Similarly to commonplace framings of metadata (see e.g., Mayernik, Reference Mayernik2020; Tennis, Reference Tennis2008), paradata is in some settings data that is always created to function precisely as paradata. In other settings, it – like provenance data – is a ‘by-product’ of the activities of individuals, groups or institutions. Paradata also stands in other, more complex relationships to provenance data and metadata, and the intersections and similarities between the concepts are many. The significance of paradata and how it relates to metadata or provenance data has to be determined on a case-by-case basis, although there are some general patterns to how the concepts are employed together, as previously discussed in this chapter.

The way of thinking about paradata in a more tangible sense also points to paradata challenges that are important to discuss. In this view, both the definitional and the use case facets of paradata are understood to be bound to the practices and regimens of their setting of application. That being said, data like paradata always travel across several settings of application, where practices may contrast or clash (Brown and Duguid, Reference Brown and Duguid1996; Star and Griesemer, Reference Star and Griesemer1989). In relation to paradata, this might mean that there are gaps between the understandings and conceptualisations of paradata that risk causing paradata use becoming difficult. Such gaps can manifest also in varying levels of compatibility in information system design and data ontologies with particular paradata-related tasks, purposes and understandings. Examples include research communities with more or less matched comprehensions of paradata, and data models that render paradata in ways that do not match what is required for success in a particular use case (cf. Hansson and Dahlgren, Reference Hansson and Dahlgren2021; Kim, Reference Kim2021; Mayernik, Reference Mayernik2011).

Related to this point are a range of often-present insufficiencies connected to the concept of paradata. One insufficiency is that paradata never is a ‘reality-checking’ implementation (Beacham, Reference Beacham2011, p. 51). No amount of paradata or combination of paradata can give complete access to data-interaction histories, and the work of, for example, reusing data is surely facilitated by paradata, but never fully driven and accomplished by paradata. Tasks involving the use of paradata to achieve some aim or result must always be supplemented also by other non-paradata resources and efforts.

Another insufficiency is that paradata itself is a data type that requires interpretation and sufficient interpretative conditions. This could be taken to mean that, theoretically, there would be reason to to supplement the paradata with additional paradata describing how and why it was created and documented (cf. Gant and Reilly, Reference Gant and Reilly2018; Huggett, Reference Huggett2020; Li and Sugimoto, Reference Li and Sugimoto2014). This ‘paradox … of potentially infinite regression’ (Huggett, Reference Huggett2020, p. 3) must, again, be considered in relation to practice where case-specific objectives and requirements of use, together with available resources to harness or create paradata, will be the actual factors limiting and shaping what interactions with paradata are relevant and possible.

The many definitions of and approaches to paradata are, in some respects, an obstacle for grasping and employing the concept. Equally, from a different viewpoint, it might not be considered desirable to lessen the complexities associated with paradata. As discussed later in this volume in Chapters 4–6, standards and other meta-framing attempts can be useful, but so also can the diversity of meanings and approaches to using paradata for various ends and means throughout paradata’s many fields and scenarios of use. As observed by Furner (Reference Furner2020) in his discussion of the wealth of available metadata standards, it is important that connectivities exist between communities of concept use. Such connectivities can be built on the basis of clear and well-documented applications of the concept that make visible connections and disconnects across different scenarios of its application, as has been attempted in this chapter and which this volume pursues further in the following chapters.