The Timescapes Archive: a stakeholder approach to archiving qualitative longitudinal data

I say ‘metadata’; you say ‘context’

One of the first requirements of integrating archiving with research was the need to address the archival obligation to document data early in the design phase of the research. There is growing recognition that managing data, including creating data documentation, early in the research process facilitates data sharing. Broadly speaking, researchers and archivists agreed on the value of data documentation and the need to collect it. There was also shared understanding that documentation included items such as: project descriptions, methodologies, interview documents, field notes giving insights about participants, substantive themes, and the conditions under which data were generated, data descriptions, conditions of consent, confidentiality, and so on. Beyond this general level of agreement, important differences arose, and differences in terminology had to be addressed and resolved.

Within the broad category of documentation, archivists distinguish a subset, called metadata, or data about data: ‘Metadata are a subset of core data documentation, which provides standardized structured information . . . typically used for resource discovery . . .’ (UK Data Archive, 2010a). Examples of metadata include data creator’s name, date of data creation, and keywords that describe the data. Timescapes had to confront a specific challenge here: although some structured metadata existed, we were seeking to develop a more comprehensive set of metadata particularly adapted for QL research. Most social science data archives follow a standard called the Data Documentation Initiative (DDI, 2010). Within the DDI general categories exist such as study title and creator, but some of the specific fields most useful for QL data, such as life course and temporal terms are not included. The difficulty arose in that archivists were, in a sense, asking researchers to collect metadata while the precise list of what to collect had not yet been specified. Early conversations between archivists and researchers used general definitions, preferring to err on the side of inclusiveness rather than omit key information. One early project meeting defined metadata as ‘fieldwork/research materials, historical/cultural/locality source materials to fully contextualize the research’. This was not an archival definition of metadata, but it was inclusive, with the added benefit of being comprehensible to the researchers.

Over the course of repeated conversations, the word ‘context’ became a bridging term. When researchers asked archivists to be more specific about metadata, the archivists asked for anything that would ‘contextualize’ the data, a term that had meaning within each domain, though it meant relying at this stage on a quite general definition of ‘context’. For qualitative researchers, context is linked with a relational understanding of data, that is, the impossibility of separating ‘data’ from ‘context’. Context, qualitatively speaking, is more than a static set of variables that surround persons, actions or situations but a fluid constellation of contingent factors, the salience of which emerge through everyday interaction and interpretive processes (Holstein and Gubrium, 2004). Data can only be said to be comprehensible within their context, ranging from interpreting an interview extract by situating it within the larger narrative to which it belongs, or analysing a turning point in the course of a life only by seeing it as part of the full biography in which it occurred and in relation to other biographical accounts. The archival sense is usually more narrow and technically defined, referring to the variety of documentation that might accompany data (UK Data Archive, 2010b). However, the flexibility of meanings permitted all groups to concur on next steps.

There was agreement to collect detailed metadata, and agreement that researchers were the custodians of richly detailed descriptive metadata. The specific problems were: what to collect and in what formats? Alongside detailed notes about fieldwork encounters, we settled on three categories of descriptive metadata. First, identifying information was to be documented on the interview transcription cover pages, showing the name of the project, depositor, interviewer, date of the interview and information on the interviewee and setting. This page was modelled on guidelines provided by the British Library Sound Archive Oral History Section and a transcription template from the UK Data Archive.

Second, we collectively defined a list of keywords that could be used to identify the thematic content of each data file. This list comprised 20 conceptual and descriptive terms, such as Childhood, Youth, Adulthood, Past and Future, that could be assigned to all interviews. The list was not exhaustive; researchers could also assign their own keywords to indicate specific thematic content of a data file. But the list provided a controlled vocabulary of basic terms to aid thematic searching and retrieval of data across the dataset. Third, we defined a structured set of what we called ‘base data’. These include socio-demographic characteristics of the participants, e.g. date of birth, living arrangement, employment (or employment of parents), income, and so on. These items were collected in part to allow comparative work to be undertaken with quantitative longitudinal studies such as the British Household Panel Survey and its successor, Understanding Society. With agreement on these three categories of metadata, the archiving team could develop the schema required to ingest the data into the archive and to enable QL-specific resource discovery of the archived materials.

Our differences were not ‘just’ linguistic ones; different languages reflected substantive underlying debates about the relationships among data, context, and documentation. Within archival practice, data, documentation and context are distinct and separable. Indeed the concept of long-term preservation states that documentation enables data to be retained in a way that renders it ‘independently understandable’, defined as:

A characteristic of information that has sufficient documentation to allow the information to be understood and used by the Designated Community without having to resort to special resources not widely available, including named individuals. (OAIS, 2010)

Data are different from their documentation, and documentation should render data understandable without the need for additional context or information from individuals such as the data creators.

This archival view, which highlights the stand-alone nature of data, stands in contrast to the relational, contextual and embedded view of data held by researchers as described above. Some have argued that these views are fundamentally opposed. Mauthner and Perry (2009), for example, claim that archiving presupposes a foundationalist epistemology. In practice, however, the issues did not seem to lend themselves to such simple, binary oppositions. Most Timescapes researchers self-identify as working within epistemological frameworks of constructionism, with data and context of a piece. Yet all agreed to contribute data to an archive, not without some reservations, but clearly without subscribing to a foundationalist stance. This cannot be seen as cognitive dissonance, but rather, a layered, situated appreciation for the complexities of the archiving project and a growing interest in exploring and contributing to the mediation of knowledge in the archive.

Timescapes researchers shared the view that there is unique and ‘uncapturable’ information constructed during direct contacts with participants. However, appreciating the uniqueness of these encounters did not devalue the process of richly documenting how the original data were gathered and under what conditions. The researchers recognized two realities. First, they wanted to capture and record comprehensive documentation because they had the most detailed knowledge of the field and wanted later users to benefit from their knowledge. Second, no matter how thorough their efforts, everyone also knew documentation of this context would always be partial. This did not represent settling for a ‘half-done’ job but rather it demonstrated an ability to move flexibly between subjective and objective relationships with data. Team discussions also showed that we understood that future researchers would, regardless of our work, re-contextualize the data, perhaps using materials we had provided, but also bringing their own critical and reflexive views to bear on the data (Bishop, 2007; Moore, 2007).

Hadfield (2010) provides a sensitive accounting of this nuanced way of proceeding within the Motherhood projects. The team placed great emphasis on detailed and rich description of the research encounter, settings, interviewees, and also the feelings and responses of the researchers. After team discussion, it was deemed that these raw notes were too revealing to deposit in the archive. Also, the anonymization that would have been necessary in some cases removed significant contextual information from the data. The team decided, therefore, to create ‘parallel’ reflexive field notes that were written from a self-aware stance and with the expectation that they would be archived. Also, notes were added to transcripts alerting future readers to the loss of significant information from anonymization. With the mediating influence of the archive, researchers were encouraged to pro-actively build the probability of QL data re-use into their primary research practices and data management, thereby allowing for an expansion of the analytic lens to encompass secondary use. Allowing for broad understandings of context – both technical and epistemological – to enrich the metadata was crucial to this process.

Negotiating standards and flexibility

‘The primary objective of all archives is to select, preserve and make available for use documents or information which have permanent or continuing value’ (UK Data Archive, 2010c). To achieve this objective, archives must satisfy requirements of extensive and complex legal and regulatory frameworks. For data archives, these include copyright laws, the Data Protection Act 1998, and other national and international directives.

Researchers, too, operate in their own complex environments, made up of university administrations, professional bodies, ethical guidelines, funders’ conditions, research cultures and so on. For empirical researchers, a central objective is the welfare of research participants and ensuring that they come to no harm from engaging with research (Hadfield, 2010; Spicker, 2007). This duty of care was important to the Timescapes team, particularly around issues of confidentiality, the ethical representation of accounts and interpretations, whether consent for archiving was informed where the specifics of re-use were unknowable, and whether secondary use would render participants vulnerable to exploitation or their views being misrepresented. The duty of care extends to the researchers themselves, with concern, for example, over whether the sharing of data may expose them to unfair criticism (Neale and Bishop, 2010, 2011). In addition to a primary emphasis on care, most researchers subscribed to an ethical stance that emphasizes the particularities of any ethical question, and typically calls for negotiation in resolving ethical questions, rather than relying on the application of general rules (Mauthner and Edwards, 2010).

Separating out these central objectives above should not be taken to mean that either archivists or researchers were insensitive to the concerns of the other group. The archivists were familiar with research practices and their ethical needs; the researchers were committed to the protection and long-term preservation of the valuable data they were collecting. However, the distinctive priorities of the two groups shaped how key practices evolved during the project. Only through extensive negotiation could the objectives of both groups be reconciled. Examples are discussed across the three strategies used to make data shareable: gaining informed consent for archiving, altering data to protect identities (e.g. anonymization) and regulating access to data through a system of access controls.

Informed consent

Informed consent is an ethical requirement for most research. Timescapes researchers and archivists shared strong commitments to the ethical use of data, but their primary roles shaped the particulars of how these commitments were perceived and acted upon (Bishop, 2009). The Timescapes archivists sought to implement consent procedures early so that the data collected would be sharable. The archivists started with the model consent form developed at the UK Data Archive which reflected legal requirements of data archiving: the preference for written consent, the timing of when consent should be sought, and some provisions of the consent form itself.

The Data Protection Act regulates how data are handled, specifically personal or sensitive data. According to the Data Protection Act, personal information is about a living individual that identifies that individual. Sensitive data are information such as racial or ethnic origin, religion, medical information, etc. (Data Protection Act, 1998). If research data to be archived contain personal or sensitive data about informants, consent is needed from each informant for such data to be processed, ideally in writing. Formally, this applies only to personal data. Therefore, if a person cannot be identified – because identifiers have been removed or if the data have been fully anonymized – then the Data Protection Act no longer applies. For much qualitative data, it can be difficult to assure absolute anonymization, and thus the safest (legal) stance from an archive’s perspective is to have written consent in place.

Archives also tend to recommend that consent for all purposes (research, publication and archiving) be gained as early as possible. While not a legal requirement, this advice emerges from seeing many cases where such consent was not obtained, and for reasons such as inadequate funds or loss of contact with participants, consent for archiving was subsequently impossible to obtain, rendering potentially valuable data unshareable.

Another factor for archives to consider is that of copyright. The UK Data Archive recommends that researchers ask their participants to transfer copyright to the researcher. This protects the researchers by enabling them to legally authorise the UK Data Archive to make copies of digital materials for purposes of digital preservation. The net effect of these requirements is that the UK Data Archive recommends a written consent form, that consent be sought early in the research, and the form explicitly asks participants to transfer copyright to researchers.

While Timescapes researchers shared a commitment to the ethics of consent, they had concerns about the first version of the consent form. For example, some potential participants came from very poor households, and had had repeated and negative experiences with bureaucratic forms and legal documents. Researchers were also justifiably worried that the language of the form was intimidating and that a less formal, verbal approach would work better. Furthermore, researchers made a case for separating consent for participation in the research from consent for archiving. They wanted to gain optimal levels of participation (and some projects had extremely challenging recruiting profiles to fulfil) and build rapport with participants before discussing archiving. In the latter stages of research, they argued, participants would have a much clearer understanding of the data they had produced, and would be more able to grant authentic informed consent. The longitudinal method meant there would be repeat contacts with participants and numerous opportunities to seek consent later in the process.

The process of producing a consent form acceptable to all the projects took a great deal of time, including drafting the form, holding consultations with representatives from each of the seven projects, incorporating their feedback into the form and keeping the wording of the Timescapes form aligned with changes and updates in UK Data Archive policies. The effort yielded a standardized consent form that covers areas of consent for participation, research outputs and data sharing and archiving. The outcome has been positive: all the teams agreed to use the form and the main changes were to customize the form, while retaining its main meanings. For example, customized language was needed to adapt the form to certain groups, such as young people. The main deviations were that rather than written consent, some projects gained verbal consent and recorded and transcribed the consent agreement, while others deferred asking for consent for archiving until later waves of fieldwork were conducted. Finally, there was good collaboration in producing information sheets that accompanied consent forms. The archivists drafted these scripts to help researchers explain the idea of a data archive to participants. This boilerplate text explained issues such as how digital data is shared (making clear that this does not occur publicly on the web); the ways data can be re-used; reasons for needing copyright; and so on. Researchers then customized this text into their own project-specific information sheets.

Anonymization of data

A similar procedure for collaboration was followed around producing guidelines for anonymizing data. This is seen as the key mechanism for protecting the identity of research participants. Again, draft guidelines were available from the UK Data Archive and from the Making the Long View archiving project. Exchanges took place between archivists and researchers, but the challenges in this case were greater in a number of ways. First, the initial guidelines proposed proved to be too complex and even when simpler ones were produced, not all projects were aware of the changes and some confusion ensued over which version was in effect. More fundamentally, there are widely varying views on appropriate levels of anonymization, from none, to light touch, to comprehensive. The oral historians within Timescapes, for example, prioritize the authenticity of historical accounts, which militates against anonymization, whereas the tendency within sociology has traditionally been to comprehensively anonymize to protect the identities of these they study. It was nearly impossible to specify in the abstract what should be used. Archivists and researchers attempted to work toward the best solution for each project. Very possibly, more project-specific support from the archivists would have helped at this phase.

Access controls

In the end, the challenges of anonymization were, in a sense, partially addressed by the establishment of a system of levels of access to the data. This is an alternative route to meeting the requirements of ethical protection, one which controls or limits access to the archived data. This was seen as a viable option for researchers who were concerned about sensitive data or dangers of ‘over-anonymizing’. Across the team we negotiated four levels of access to Timescapes data: public access, registered access, approved access and embargoed. For the first level, metadata and selected screened data are publicly available and enable browsers to sample data sources and gain a feel for what the archive contains. The second level, registered use, is the primary form of access to the archive, reserved for bone fide researchers from authorized academic, policy related or public institutions. Registered users sign a licence agreeing to strict terms and conditions. The third level, approved use, enables registered users to gain access to specific data that are sensitive or difficult to anonymize. They are required to submit a further application, which is screened and considered for approval by the originating team of researchers who deposited the data. Finally, some data are embargoed for varying periods of time. The ability to grant or deny access by individual users to individual files (such as interviews) was reassuring to the researchers. This was vital in enabling some of the most sensitive data to be deposited, data that in likelihood would not have been deposited at all without these flexible controls. The process of bringing the originating team into communication with potential secondary users is also likely to be potentially productive for both.

Communication and negotiation – necessary elements in the stakeholder model

As the examples above show, communication and negotiation underpinned the stakeholder model of archiving used to build the Timescapes resource. It may seem obvious that communication would be important in a stakeholder model. Yet it is surprising how often it is neglected. It is easier to presume shared meanings than to probe: do we all mean the same thing here? And not focusing on differences can be useful for a time – the multiple definitions on context, for example – yet eventually meanings had to be clarified. Similarly, negotiation would seem obvious, and it is, but too often only in the abstract. When it comes to allocating project resources to the travel and meetings needed for the processes, this is more challenging. Timescapes made good efforts in these directions, yet everyone believes that even more was needed.

Perhaps what is most salient about the stakeholder model is that it was a deliberate choice; that alternative paths could have been taken. One possibility could have been to impose a far more structured, hierarchical, directive scheme. That is, the archiving team could have attempted, for example, to mandate the use of a standardized consent form, anonymization procedures, and so on, without consultation or negotiation. Certainly much time would have been saved. However, despite any shortcomings of the stakeholder approach, the Timescapes team continues to believe that this alternative would have been much less acceptable, and thus less effective. Imposing this kind of control would have been counter-productive in an initiative designed to respect and encourage a grounded and flexible approach to fieldwork and data generation across our network of projects. Even if it were possible, it is antithetical to the collegial culture of research and of this research team and indeed to most academic enterprises. Finally, it would not have been a proper foundation for the kinds of collaborative work Timescapes intended and, indeed, has fostered.