Mobile researchers,immobile data: Managing data (producers)

Data curation, mobility, and reuse

The term ‘research data’ can include a vast variety of objects – numbers in datasets, instrument readings, photographs, transcribed text, archival materials, observations, field notes – all of which are by their expected evidential value (Mayernik, 2019). Early laboratory studies (e.g. Lynch, 1985) pointed out that the recognition that an observation, a measurement, or a document constitutes data is already an act of interpretation (Borgman, 2012, p. 1061). Data are not, as the word’s etymology suggests, ‘given’; they are complex achievements that pertain to particular stages of the research process. Lynch (1985) dwells extensively on the work involved in creating typical representations and in deciding on the nature of concrete pieces of evidence. This corresponds to the empirical difficulty of separating data from the practices of their production. Researchers interviewed for this study frequently regarded both the physical samples they collected as well as the outcomes of some analysis procedure as ‘data’, as products of scientific work pertaining to different stages of the research process. In that sense, managing data falls in line with managing the stages of the research process (as is evident in attempts to explicate data handling by reference to the research data lifecycle, see Mosconi et al., 2019).

Contemporary scholarship has interpreted data sharing as a practice that enhances the mobility of data to enable unspecified, global data reuse and to improve scientific rigour and reproducibility of results (Leonelli, 2020). Accordingly, this literature uses the term research data management (RDM) to describe specific policies and procedures geared towards making data more shareable, not simply any practice of managing research data (a usage I adopt in this paper). Even delimited in this way, there is a vast literature studying data sharing (Tenopir et al., 2011, 2015; Unal et al., 2019), documenting a staggering variability in practices, infrastructures, approaches (Akers & Doty, 2013), and data types (Borghi & Van Gulick, 2018; Borgman, 2015; Whitmire et al., 2015). Recent work on data mobility (Leonelli, 2020) is indebted to earlier scholarship on laboratory practices in describing data as the results of manifold processes of construction (Knorr Cetina, 1991, 1999, 2007) and in pointing out the tremendous amounts of work and curation necessary to make data mobile (Mirowski, 2018). An important corollary to this view (Leonelli, 2016) stresses the emerging role of data curators in explicating the tacit aspects of the research process to extract data from their contexts of origin. Data curators ensure reusability through describing ‘techniques, assumptions, methods, materials, and conditions under which data are generated’ (p. 94).

In this literature, the social role of data curators is largely explained as maximizing the evidential value of data by helping to expand the range of contexts of reuse (Leonelli, 2016, p. 194), thereby echoing Collins’ (and others’) understanding of tacit knowledge (Collins, 1985). However, a significant portion of the literature (Mayernik et al., 2011; Wallis et al., 2013) takes for granted that data sharing is beneficial for the conduct of research. This may be because there are relatively few in-depth case studies of how data sharing benefits research practice, and these studies do not situate these activities in the context of broader transformations of academia occurring alongside the Open Science transition.

Academic capitalism and academic mobility

Since the 1980s, universities are increasingly incentivized to capitalize on the knowledge they produce (Etzkowitz, 1998). Activities such as university start-ups, university patenting, technology transfer, and university-industry co-authorship have increased significantly (Shibayama, 2012). Through changing the requirements for public research support, governments hoped to use universities ‘as surrogate agents for industrial policy programs that governments were unwilling to undertake more directly’ (Etzkowitz & Webster, 1994, p. 497). Universities increasingly act as producers of knowledge and human resources (Etzkowitz & Leydesdorff, 2000, p. 110). Accounts of ‘academic capitalism’ describe the complex social processes of marketization, increasing ‘managerial’ governance, and increasing competition within and between universities (Clark, 1998; Münch, 2014; Rhoades et al., 2019; Slaughter & Rhoades, 2010), which has led to a loss of autonomy and strong uniformity in terms of research trajectories and institutional forms (Münch, 2014).

At the same time, academia has become an increasingly globalized job market. Transnational academic mobility has gained in importance (Kim, 2017; Morley et al., 2018), strategically promoted by policy actors as a criterion for career evaluation. The importance of being mobile has overwhelmingly been framed in epistemological terms (Sautier, 2021), stressing mobility as a positive force for researchers and institutions alike. Within this logic, universities preferably employ senior staff who obtained their qualification elsewhere because ‘studying abroad is believed to broaden networks and prevent parochial cultures’ (Bennion & Locke, 2010, p. 11). However, the development of academic entrepreneurialism has also meant that an increasing proportion of the academic workforce experiences precarity, particularly during and after completing their PhD (Butler-Rees & Robinson, 2020). Job market conditions for advanced degree holders have become increasingly insecure as the availability of permanent academic positions declined (Acker & Haque, 2015, 2017).

While geographical mobility of academics is not a new phenomenon (Fernández-Zubieta et al., 2015), there has been a significant increase in the past decades following the internationalisation and massification of higher education (Sautier, 2021). In the process, mobility has come to be regarded as a sign of research excellence (Sautier, 2021) and a tool for increasing academic collaboration. Indeed, a European academic labour market has been defined as the primary goal of the European Research Area (Ryan, 2015; Ulnicane, 2015). Academic mobility means the circulation of researchers as well as the prescriptive idea that advancing one’s career is conditional upon ‘going abroad’ (Appelt et al., 2015). While the political interpretation of mobility has been largely positive, some commentators have also expressed concern for what has been described as a romanticised version of mobility (Robertson, 2010).

The Austrian academic job market exhibits some degree of internationalization, along with low levels of job security over an extended period. Like Germany and Switzerland, academics face long periods of insecurity after finishing their PhDs. The resulting conditions have been described as a ‘survivor model’ (Enders & Musselin, 2008), where aspiring academics go through years of trial and insecurity before (possibly) obtaining a permanent position or leave the field involuntarily. As in other countries with similar structures, the availability of tenured positions has declined significantly over the past three decades, with increasing competition over fewer permanent positions (Sautier, 2021). In 2002, a new Universities Law (UG2002) further institutionalized employment insecurity for non-tenured staff by turning universities into self-regulated bodies under common law (until then, tenured university staff had been appointed as federal employees) and making consecutive fixed-term employment illegal. While the aim of the law was to protect non-tenured academic staff, universities interpret it differently and routinely terminate non-permanent contracts. An amendment passed in December 2020 has reinforced this practice by limiting total contract duration at the same institution to a maximum of eight years. This has meant that employment below the level of associate or full professor has become much more flexible, allowing for employment in multiple and overlapping projects (at least in theory). Both internationalization and casualization of academic careers extensively affect the professional and personal development of early career researchers during a biographical period that involves multiple (often competing) commitments such as ‘improving one’s academic CV, applying for stable jobs, establishing a relationship or having a child’ (Sautier, 2021, p. 804).

In what follows, I do not discuss the normative readings of academic or data mobility any further. Instead, I investigate the relationship between the arguments for allowing academics and data to move freely, and how that can be achieved. The mechanisms are different in each case, but the arguments both appeal to broadly epistemic concerns (see Sautier, 2021 for academic mobility and Borgman, 2012 for data mobility). In what follows, I am indebted to the contention, central to Grounded Theory (Glaser & Strauss, 2009) but found already in the authors’ earlier work on dying (Glaser & Strauss, 1972), that analysts should strive to understand which problem social actors are trying to solve with a specific action. I develop the view that data handovers solve a specific problem – a problem that is different from what open research practices aim to achieve.

Neo-institutionalist perspectives (Dobbin & Sutton, 1998; Meyer, 2008, 2010) suggest that while early adopters of open practices (the life sciences, among others) have done so based on practical necessity (in terms of improving certain forms of doing research; Leonelli, 2012; Thessen & Patterson, 2011), later adoptions occur (mainly) based on perceived legitimacy (Dobbin & Sutton, 1998), following ‘definitions, principles and purposes that are cognitively constructed in similar ways throughout the world’ (Boli & Thomas, cited in Drori, 2008, p. 460). For data sharing, the relevant principles are those of transparency and reproducibility, with the (implicit) expectation that (all) research fields will take a similar trajectory of openness by adopting a specific set of practices. In what follows, I develop counterexamples to this view. I argue that man of these instances are better conceptualizes as data handovers; that is, local forms of data release for a circumscribed audience. This suggests that functionalist explanations of open data (as solving specific research problems) are only viable in certain research contexts and stages of development.

Research data management in Austria

While the recent trend towards Open Science spawned renewed interest in data practices (Leonelli, 2016), this work is predominantly built upon empirical insight from Anglo-American academia. However, some of the conditions, such as the structures research funding and employment contracts, may not apply elsewhere. Over the past decade, many countries have started to develop and implement Open Science policies (Lilja, 2020; Olesk et al., 2019). Austria is a particularly interesting case as it has seen a number of initiatives since the early 2000s (Mayer et al., 2020). The Austrian Science Fund (2018) has been promoting Open Access since 2004 and has been a member of cOAlition S (2022) since 2018. The Open Science Network Austria was founded in 2012 as a think tank to facilitate the implementation of Open Science. In 2016, the group published its first recommendations for a national Open Science strategy (OANA/UNIKO 2016). The 12 ‘Vienna Principles’ on the future of scientific publishing from the same year (Kraker et al., 2016) have served as a framework for strategy and policy development, both nationally and internationally. Since then, many Austrian research institutions have developed Open Science policies, and some have started to implement institutional repositories. As of 2019, Open Science is part of the HRSM (Structural Funds for Higher Education) focus on digitisation, which is currently funding several projects to develop Open Science infrastructures and coordination between Austrian beneficiaries (Mayer et al., 2020, p. 18 ff.). Requirements to share data produced at Austrian universities now come either from these institutions directly or from national and international research funders.

The material presented in this paper was collected in 2019, at a decisive moment in the process of introducing RDM policies and practices in Austria. These interventions to introduce RDM were not received favourably. To make sense of this situation, I interviewed 18 researchers in science and engineering departments at one Austrian university about their data practices. Interviewees were predominantly research group leaders, though some were accompanied by PhDs or postdocs. All were aware that the research was intended to understand reactions to the policies outlined above. Their responses suggest that it is not only (nor even predominantly) concerns for global data reuse or transparency that drive their adoption of more transparent data practices. Respondents’ concerns pointed in a different direction: reuse of data within their research group in the face of increasing academic mobility and employee turnover.

As the approach is based on a sample from a circumscribed context, the findings should not be extrapolated uncritically, for instance to other national contexts or scientific disciplines. However, unlike retrospective analysis of fields in which broad data sharing practices have already been implemented, these interviews illuminate a decisive moment prior to the implementation of specific RDM policies. As the sample was diverse, with interviewees working in the life sciences, physics, mathematics, as well as various engineering disciplines (notably, no representatives of the social sciences or humanities were interviewed though), the interview results suggest that data handovers occur somewhat irrespective of discipline.

Data handovers and academic precarity

One surprising finding from these interviews is that when discussing their motivations for data sharing practices, interviewees often referred not to unbounded publication of their data but to the process of handing over data when a researcher leaves a group. This is a crucial moment in a project’s lifecycle: if a project is to outlast any individual’s tenure in a research group, researchers must leave behind accessible and useable data when they leave. Because PhD students and postdocs perform the bulk of data collection and analysis, such departures are a regular occurrence: ‘Of course, a lot of the data come from studies by PhDs or postdocs, work that ends with a PhD or master thesis […]. Our record is, maybe not a database, but scientific outputs, our publications’ (Associate Professor, Mechanical Engineering). The pressure on researchers to manage data is particularly high during the final phase of a PhD:

[Students] are under a lot of pressure and stress, especially during the final phase of the dissertation or master thesis, [a fact that] tends to be forgotten, and then we are pressured to provide a grade, and then people leave. I have thought about the potential to optimize, because we have lots of data and it works, but it could be improved. (Senior scientist, Technical Chemistry and Process Engineering) ¹

This final phase is a challenging moment in data curation because this is usually when data producers already have large workloads that leave less time for data curation tasks. Documentation is rarely considered research. Data management is regarded as an administrative task that predominantly benefits not the data producers, but their respective successors or those on permanent contracts:

This is the first point: [we need] long-term archiving and interoperable [data], that would be the second point. How do we go about managing all this data centrally so that every project researcher can simply add their data for those who are here long-term and do not leave after their thesis? (Full Professor, Technical Chemistry and Process Engineering)

Data management is performed in service of the research group (‘those who are here long-term’), not for the benefit of the data curator. Data handovers are thus ambivalent. There is a conflict inherent to local data release because any efforts devoted to data curation will most likely benefit the successors rather than the data producers themselves:

Okay, it’s not mine, it belongs to [the university], and that I accept. But am I really supposed to hand [data] over to my successor? I mean, I do it, but don’t like it, they should not have it too easy. (Full Professor, Mechanical Engineering)

Governance is needed to separate research data from those who perform data curation (Senior scientist, Computer Science and Biomedical Engineering):

[We need] a concerted, focussed effort to build such a database, with corresponding data entry masks, […] and someone watching over people putting in their [data] correctly and [in full]. You put in the work once so that in the long-term [RDM] will be […] less time-consuming. (Full Professor, Technical Chemistry and Process Engineering)

Research data management, such as the standardisation of (meta)data practices, is interpreted as a solution to the problem of non-tenured researchers leaving research groups, but how this is implemented varies. Few institutes have centralized RDM; PhD students are expected to manage their own data without receiving much guidance, which entails that supervisors develop processes for students to manage and hand over data. Respondents in managerial roles (e.g. deans, heads of departments) think about RDM issues primarily in terms of benefits for administration, and only secondarily in terms of benefits for the conduct of research.

‘There’s not a lot of demand for me to do that’: Motivations for adopting data sharing

Motivations for or against sharing data have been discussed extensively in the literature (e.g. in Ceci, 2018; Houtkoop et al., 2018). In line with previous studies (Bechtel, 2020; Boumans & Leonelli, 2020; Cambrosio et al., 2020), some interviewees did portray data management practices as driven by epistemic considerations and the desire to foster the development of a research field:

This was sort of an early necessity, because only when… you have the database from all [bacteria stems] are you able to compare them and then you can really find out the unknowns, whereas if you only have your own data, the certainty is much, much lower, and those were very smart people in microbial ecology [which] was always a community where these things were discussed. (Full Professor, Technical Chemistry, Chemical and Process Engineering, Biotechnology)

The adoption of data sharing enabled insights that would have been impossible without data integration. However, data integration is more of a concern in some fields (especially the life sciences; see Bowker, 2000) than in others. While researchers in all fields need to address the problem of data handovers, not all aim to make data globally available for integration:

In terms of availability, making truly large amounts of data globally available is not, there’s not a lot of demand for me to do that. I think the question is, what is available that will help us […] it’s not customary in our community to really make [raw] data freely available. (Associate Professor, Civil Engineering)

This is not to say that interviewees did not see any personal benefits from engaging in data sharing. In the view of many respondents, data sharing imposes discipline on researchers to develop data management structures that enable proper reuse and data security: ‘If there is a uniform data structure for one area at least, where you can say, okay, now I can find my colleague’s doctoral candidate’s data easily, that is a good idea’ (Associate Professor, Civil Engineering Sciences). The personal discipline associated with data sharing are interpreted as conducive to good scientific practice, in the sense of forcing researchers to keep datasets well organized. On the other hand, this discipline is precisely what some perceive as an additional ‘administrative’ burden.

Reinterpreting data management: Managing data (producers)

Another surprising finding was that respondents see effective data management not as intrinsic to a particular research protocol, but as emerging from the social connections between successive data producers. The FAIR guiding principles for research data management, for example, construe open data in terms of process and protocol: they mandate that data be findable, accessible, interoperable, and reusable, and suggest using DOIs, standardized communications protocols, standardized vocabularies, and metadata to achieve these aims (Wilkinson et al., 2016). As has been discussed extensively in ethnographies of non-Western research contexts (Rappert & Bezuidenhout, 2016), however, this interpretation is euro-centric, in addition to being narrowly geared towards specific kinds of (digital) data. In interviews, respondents acknowledged the values of findability, accessibility, interoperability, and reusability, but fostered them not through managing data but through managing data producers. For instance, one supervisor remains in touch with lab alumni explicitly to ensure data interoperability and reproducibility:

You can save the physical data for all eternity, with the guarantee of being preserved forever, but I don’t think highly of that… it’s highly inefficient when my young colleagues work on their own and develop [their own ways of managing data]. If I have the chance of bringing them into contact with their predecessor for a half-day or a day, this is one hundred times more effective than doing it alone. (Associate Professor, Mechanical Engineering)

Likewise, findability is not necessarily a result of appropriate data curation, but rather of managing data producers. This is especially evident when non-digital data are concerned:

[Lab notebooks] are archived, yes.… None of my students have finished their theses yet, but this is exactly where they are collected. When the notebooks are done nicely then [results] can be reproduced. That does work, and like I said, the data are handed over to the supervisor, in the form of USB sticks, say, and the supervisor takes care of them until publications are ready. (Senior Scientist, Technical Chemistry and Process Engineering)

(Analogue) lab notebooks provide a good example of what is at stake here, as the administrative aspects of data handovers are less hidden away by technology. Issues of data management are not restricted to digital technologies or specific data types. Since findability is an organisational achievement, reproducibility and replicability depend on how well the succession of PhD candidates is managed:

[Reproducibility] is a big problem for us, but this is down to people’s know-how… it is not at all easy to pass on the know-how from one PhD candidate to the next, because when one leaves you have a 6-month-break before the next one starts. (Full Professor, Electrical Engineering)

Failures in reproducing results are—at least in this instance—attributed to the break that ensues between two successive PhD candidates. Indeed, some informants trust personal contacts over technical solutions: ‘The best data management works via a quick phone call […] which trumps any documentation’ (Associate Professor, Mechanical Engineering and Economic Sciences). Whether ‘a quick phone call’ is appropriate will depend on the data in question, but the excerpt certainly documents a tendency to trust social arrangements over technical solutions. Data release thereby turns into the problem of managing successive data producers.

Consequently, interviewees saw data curation as a matter of managing data producers, and not in terms of ensuring data reuse. This happens in one of two ways: either in terms of successive data producers as above, or in the sense of organising division of labour, for instance, between scientific and technical staff:

[PhD students] are employed at the university and technicians are the interface to building the model. The PhD students do the planning and help the technicians build the model, they have to see to it that it works […] we have two technicians taking care of the measurements, setting up the equipment, who set up the equipment for the […] measurements and usually do the measurements, but regrettably they are not the ones doing the data analysis […] They do the measurements and then hand the data over, which is when the PhD students come back in, as they are the ones who proceed to deliver results based on these data. (Associate Professor, Mechanical Engineering)

Even where technicians are involved in the process of data creation, data analysis and curation are done by PhD students. In part, this is because creating a unique data set is considered fundamental for receiving a PhD. That non-academic staff is involved in data creation feeds into the view that data management has less to do with good scientific practice than with administrative ‘box-ticking’:

Especially as concerns data management, it should be kept simple, because you do see this among non-scientific staff, but also with PhD candidates, they drown in work and resent additional bureaucratic expense (Associate Professor, Mechanical Engineering)

In the above excerpt, data management is squarely associated with bureaucracy, not with good scientific practice.

High-throughput universities: Managing the mobility of researchers, not data

The findings discussed above suggest that the mobility of the academic workforce has adverse effects upon the continuity of research projects. This interpretation is notably dissociated from those readings of academic mobility that have framed it in predominantly epistemological terms (Sautier, 2021), stressing its positive impact on researchers as well as institutions, for example as a panacea for academic parochialism. However, rising academic precarity has also meant that pressures to secure data from non-tenured researchers are increasing:

When data amounts increase, when projects develop in such a way that you have to archive more data, handle more data, then answers will develop how to deal with this situation […] my advice would be to train, not at the level of professors, but at the level of Post-Docs, because they are the ones who handle data. I have nothing to do with that, come to think of it […] I don’t have data, but others do […] I need to be able to trust that people manage their data well, that once PhD projects are finished, that the data are properly stored and made available. But honestly, everybody knows from experience that the connection from one person to the next never works 100%. Maybe in rare cases, but there will always be a cut when people switch [institutions]. (Full Professor, Mechanical Engineering)

In particular, respondents embrace both the normative assumption that data sharing is positive as well as the normative reading of academic mobility as enabling collaboration and individual employability. While the above excerpt acknowledges that increasing data amounts pose an issue, the last two sentences identify the root problem as handing data over from one person to the next. This situation is pressing for all researchers, but fixed-term contract researchers shoulder the bulk of data collection. Some respondents in administrative roles deem high fluctuation desirable because new staff is believed to bring in new ideas, expressly making a high percentage of (fixed term) PhD positions part of their strategy. However, respondents not in administrative roles reported that employing researchers on fixed-term contracts is associated with problems of knowledge loss:

Our doctoral quota is high […] two theses per year, which means that I have a lot of fluctuation among my assistants and project assistants, which makes research data management a very pressing problem… The biggest problem is that people leave with all their knowledge. (Full Professor, Mechanical Engineering)

Research groups require solutions to deal with the effects of academic mobility, creating the problem of how to manage data handovers: ‘The successor should start before the first one leaves, and in most cases we manage to do that’ (Associate Professor, Mechanical Engineering and Economic Sciences). Where the desired structures for data handovers are in place, RDM is only of secondary concern. This isn’t always easy, though, because securing data succession through organisational processes requires adequate funds:

[Staff turnover] is quite annoying, really, and my only way around that is good documentation and, if possible, to have people’s contracts overlap. But this is often impossible due to the financial situation. (Associate Professor, Mechanical Engineering)

The problem of making data available for reuse seems to occur predominantly for research that is publicly funded (whether basic or applied; there are exceptions, such as research on issues of security). Recall here that securing the availability of publicly funded research results constitutes one fundamental rationale for global data release (Borgman, 2012). The same is not necessarily true for data from privately funded research, because in those instances, funders often have an interest in keeping data confidential. Additionally, the short-lived, project-based nature of industry-funded research means that the structural problem of handing data over does not seem to be as pressing in those instances:

Those research topics are partitioned into small project chunks which means that we often succeed in completing a project within the lifespan of one PhD thesis […] There, the problem of [data] handovers is not as pronounced. (Associate Professor, Mechanical Engineering)

This solution to the problem of leaving data behind is not technical or social, but rather turns on project scope. Whether data handovers become problematic or not therefore depends on external circumstances such as project duration, and whether it is possible to answer a research question within the time frame of a research project.