Rubbing against data infrastructure(s): methodological explorations on working with(in) the impossibility of exteriority

ABSTRACT

The article focuses on infrastructures as heterogeneous assemblages. Our claim is that to examine something as elusive as data infrastructure calls for an epistemological and methodological approach consistent with the fluid ontology of the object of study. Moreover, we assert that there is no position of exteriority from which to critique data infrastructures. Our question is thus primarily methodological: How might we analyse the relational and heterogeneous nature of data infrastructures in a way that both problematizes and builds on the impossibility of exteriority? We examine two dispositions of data infrastructures: first comes their provisional, dispersed and entangled character, and second, the affective attachments imbued therein. Overall, this paper, first, returns to the ontological and theoretical perspectives on what data infrastructures are; second, makes a contribution to the methodological question of how to study them; and, third, develops a means of critique that do not locate the analyst outside the object of the critique, reminding us of the immanent relationality of data infrastructures and of the necessity to take these seriously in the methodological approach. The paper works with the existing literature on data infrastructures and research methods in education policy analysis, and with our own research data and first-hand experiences.

KEYWORDS:

• Data infrastructures
• listing
• assemblages
• methodology
• affects

Introduction: there is no outside

The review essay by Sellar (2015a), pivotal in the early scholarly debates on data infrastructures in education, points to the intensifying presence of data infrastructures in education and everyday academic work and proposes that researchers interested in data infrastructures be aware of the ways in which data infrastructures, even when studied as detached objects of research ‘out there’, simultaneously figure in and reconfigure the very conditions of research ‘in here’. Sellar suggests that we need new ways to relate to the proliferation of data because there is no position of exteriority from which to perform a critical analysis. Ultimately he asks what form of critique is possible from a context itself embroiled in the kind of policies and material conditions that are under scrutiny. Encouraging ‘careful attention to relations between our practices and those of others’, Sellar then argues that ‘studying data infrastructure would involve carefully examining relationships between different kinds of institutions, political agendas, scientific practices, education policies, modes of governance, measurement instruments and our own scholarly practices’ (774).

This article departs from the point of the impossibility of exteriority in research on data infrastructures – an important insight that has not as yet attracted scholarly attention. Our claim is that to examine something as elusive, open-ended and heterogeneous as data infrastructure calls for an epistemological and methodological approach consistent with the fluid and heterogeneous ontology of the object of study. Thus instead of, for instance, seeking to document an overarching logic to data infrastructures, or to examine the socio-, technical, material, discursive, semiotic or political specificities, infrastructural arrangements or effects of data infrastructures in education broadly defined, our primary goal in this paper is to destabilize the common binaries that divide the study of infrastructures into actor and context and to develop case examples of post-structural methodology ‘fit’ for the ubiquitous nature of data infrastructures.

Multiple interrelated processes contribute to the complex conditions of intensifying quantification and datafication under which much of education and academic work unfold. One is the overall quantification of ever wider and deeper aspects of personal and collective life made possible by developments in digital technologies and aspirations of connectivity, self-improvement, efficiency and curiosity. This is leading to a new world ‘based on continuous calculation at each and every point along each and every line of movement’ (Thrift 2004, 583). As Berry (2018) writes, ‘software and data shape and mediate our direct experience of political, economic and social systems through the automation of innumerable processes’. Across academia and diverse sectors of education, governing by numbers springing from policies of economic accountability and efficiency has given rise to a multitude of data collection exercises to monitor individual and institutional performance through publication counts, rankings, impact indices, collection of student feedback and many others (Piattoeva and Boden 2020). Academic work, for instance, is entangled with data infrastructures of academic performance and those enabling and inciting researchers to make their work visible and accessible both within institutions and publicly through social media. Data lose their connotation with interpretative scholarly work and become ‘the handmaidens of an audit culture’ (Denzin 2013, 355) or the object of value extraction (Couldry 2018), bringing together the notions of ‘data-intensive society’, ‘data-intensive economy’ and ‘data-intensive science’ (Berry 2018).

In this context, a sense of impending discomfort has crept up upon us in recent years as we have gone about our work on data infrastructures, pressed as we are with concerns over ‘academic delivery’, ‘impact’ and reporting to academic monitoring infrastructures as a mundane part of our own daily work. In past years, our research has led us to explore the rise and operation of assessment systems as tools of governance and governmentality across countries and in different historical eras. Our scholarly interest in the emergence of education sciences has likewise taken us to the realm of power-knowledge and the role of knowledge in general and of education sciences in particular in governing at a distance and subjectivity-making. We are growing increasingly aware of our own entanglement with intensifying data infrastructures and the changing role of our home discipline into ‘education data sciences’ (Williamson 2017). We thus first sensed the entangled nature of infrastructures and the impossibility of exteriority as our own scholarly and personal embeddedness in data infrastructures.

In the spirit of developing research through wonder driven by breakdown and bewilderment in one’s own understanding (Brinkmann 2014), we propose that stumbling on our own complicity in data infrastructures enables us, first, to return to the ontological and theoretical perspectives on what data infrastructures are; second, to make a contribution to the methodological question of how to study them; and, third, to develop other means of critique that do not locate the analyst outside the object of the critique. Here personal encounters serve as a means of de- and re-familiarization (Brinkmann 2014) with data infrastructures, reminding us of their immanent relationality and generative nature and of the necessity to take these seriously in the methodological approach.

Our question is thus primarily methodological: How might we analyse the relational and heterogeneous nature of data infrastructures in a way that both problematizes and builds on the impossibility of exteriority? We explore some dispositions (Easterling 2014) of data infrastructures on the basis of our post-structural approach and mindful of the extensive body of literature on the emergence and proliferation of data infrastructures.¹ While there are helpful typologies that summarize different elements of data infrastructure (see Kitchin 2014, 25–26), we focus on two understudied dispositions that enrich but also problematize existing typologies: first comes the provisional, dispersed and entangled character of data infrastructure, and second, the affective attachments imbued therein. Together these dispositions concretize the impossibility of exteriority in studying data infrastructures. Our account remains provisional and experimental in nature; the elements of the infrastructure that we examine below are much entangled and we cannot – and would not – claim to have mapped them exhaustively.

Listing infrastructures

If we start from the ubiquity of data infrastructures, how and where should we begin our investigation? Should we draw on the basic catalogue of possibilities for sociologists: focus on either the micro or macro aspects on infrastructures, the ‘symbolic’ or ‘material’, the ‘etic’ or ‘emic’? Instead of beginning with a given framework or container, we decided to follow Bruno Latour’s (2005, 27) injunction to begin in medias res, in the middle of things, without seeking to reduce an initial experience of the messiness – the crisscrossing, superimposition and the multiplicity of perspectives – of infrastructures to a seemingly panoramic view or ‘context’. We concede that ‘our most common experience, if we are faithful to it, tells us that there are lots of contradictory group formations, group enrolment-activity to which social scientists are obviously crucial contributors’ (29). Overall, this methodological choice bridges our conceptualization of data infrastructures as indeterminate, emerging and contingent (cf. Law and Ruppert 2013). The claim is thus that to examine something as elusive, open-ended and heterogeneous as data infrastructure calls for an epistemological and methodological approach consistent with the fluid and heterogeneous ontology of the object of study. This approach then also problematizes what can be taken as critique of data infrastructures. We find Foucault’s (1994a) notions of ‘limit attitude’ helpful here. The term refers to a form of critique that does not seek to go through a point of exteriority, a ‘beyond’ or ‘outside’, in order to establish disinterestedness and objectivity. Instead, critique uses that language and form of thought to experiment with what it is possible to think.

In this vein, we decided to engage with data infrastructures in a manner that approaches intellectual craft as a ‘practice that involves the whole person, continually drawing on past experience as it is projected into the future’ (Ingold 2011, 240 in Brinkmann 2014, 722). Refusing to relate to data as something externally given and thinking instead of data as ‘instances’ (Denzin 2001) that become data in moments of bewilderment or stumbling, we used our own experiences as a starting point of our examination, fusing personal narratives with larger cultural trajectories (Holman-Jones 2005). We related to calls to understand our own contexts, experiences and practices to bring to light the co-constitutive nature of data infrastructures and the academe (Burrows 2012). Yet in our work, we also sense how occasionally the academic infrastructures of quantified control rub against our research subjects – sometimes intervening in how we pursue the research, or invoking tenuous continuities between the field of research and academic life and our selves. These are not as detached as is habitually thought, and they appear to occasionally fold into each other in unexpected ways.

In this manner, as we point to the need to develop methodologies that better ‘fit’ the nature of the subject studied and to simultaneously acknowledge the conditions within which the research is conducted, we link the paper to the genre of qualitative researcher reflexivity. However, we are acutely aware of what Pillow (2003) calls reflexivity as narcissistic confession, catharsis, purification or transcendence. The agenda that we pursue here is that reflexivity is not a deconstructive or purifying exercise. To follow Pillow (2003), we bring diverse sets of data including data that we do not normally discuss in the article, or even consider to be data at all, to examine the frames with which we read the world. Our connection to reflexivity is primarily methodological, emphasizing the importance of understanding the relationship between ontology and epistemology, and how methodological choices (re)construct, and often tame, the muddled, non-linear, often contradictory and embodied reality (Petersen 2020), including our interpretation of the nature of data infrastructures in societies and education. This reflexivity 'tells us that whatever stories we tell, whatever it is that we seek to inscribe (…) will only ever be partial' (Law and Ruppert 2013, 233). This applies to both the infrastructures that we study and the social methods with which we study them: their assembled and performative nature and the partiality of both have to be acknowledged.

We approach data infrastructure as an assemblage, which is an ‘antistructural concept that permits the researcher to speak of emergence, heterogeneity, the decentered and the ephemeral in nonetheless ordered social life’ (Marcus and Saka 2006, 101). Indeed, social studies of data infrastructures have recently gravitated towards accentuating the heterogeneity and complexity of views and accounts deploying different representational modalities and different perspectives (Denshire and Lee 2013). These studies problematize the identification of infrastructures as an ‘it’ that can be neatly pinned down and its borders traced: ‘Whatever can be studied is always a relationship or an infinite regress of relationships. Never a “thing”’ (Bateson 1978, quoted in Star 1999, 379). Infrastructures as materially heterogeneous, distributed arrangement pattern and are patterned by relations as they pass through and order various materials, human, social and other. Where we set the boundaries of these relations – what is ‘inside’ and what will remain ‘outside’ – depends on our own questions and our own agendas.

Inspired by Petersen and Millei (2015; also Petersen 2020), we generated and then listed vignettes and dispatches from our encounters with data infrastructures in our own academic work, broadly defined, adding items that we intuitively felt belonged to the theme. In this manner, we rejected the problematic separation between data and analysis and instead perceived listing as a ‘“data+analysis” simultaneity’, where selecting items for the list already presupposes an analysis and invites further analysis (Petersen and Millei 2015, 132). While first making individual lists, we quickly realized their connected (and connecting) and performative potential – an item on one’s list prompted more examples on the list of the other, and so we went on, weaving in each other’s stories from fieldwork, research findings, everyday academic experiences and literature that belong neither to the subjective ‘inside’ nor to the objective ‘outside’ of academic research on infrastructures. We described and discussed each item on our list to unwrap its granular material, discursive and affective elements, asking how and why we see the different listed entries being connected. The final, joint list contained the items presented below as a smooth and concise summary that we recognize as both helpful and simultaneously a simplification of a phenomenon of a bewildering complexity (on complexity and simplification, see Law and Mol 2002):

• Our tenure track evaluation discussions and forms filled

• Frequently changing faculty funding distribution model tied to publications and journal rankings

• Workload allocation models and negotiations among staff and with faculty leadership involving these

• National journal classification system

• Indices deployed in funding applications and tenure evaluations

• Our own academic research on data infrastructures in Russia, United States and Finland including research findings and experiences of fieldwork, data analysis and reporting

• Academic and non-academic social media: Google Scholar, Academia.edu, Researchgate.net, Twitter, Facebook

The list contains entries from our studies on infrastructures, for instance, experiences of challenges with tracing a network of heterogeneous actors or our notes on studying how infrastructures feed on criticism. Yet the list also includes workload models, citation indices, league tables, academic identification numbers and platforms such as Academia.edu – all of which participate in the commensuration and valuation of our research and academic life. In some way, our list resembles Burrow’s (2012, 359) inventory of metrics operating on different scales but running through individual academics forming complex data assemblages in universities. We, too, listed emails that we received asking us, for example, to register our ORCID number for better tracking or forms that we completed to request funding for having our articles proofread (including the one at hand!) in which we were required to identify the expected numerical impact of the publication venue. But our list continued to grow in different directions, including more vague entries of uncomfortable and uplifting affective attachments to academic infrastructures.

The result thus mixes categories that initially do not really seem to ‘belong’ in the same plane. As such, the list is like an entry from the Chinese encyclopedia in oft-cited Jorge Luis Borges’ story (Borges, cited in Foucault 1970 xv).² Yet, instead of pruning the list by taking out entries that do not mesh with one another, we felt that our haphazard list is exactly what helps to point towards many currently understudied connections between different categories, destabilizing extant ways of thinking about ‘data’ and ‘infrastructure’ and cultivating a more ‘messy’ social science.

The listing exercise started with a spontaneous recounting of an episode in which data infrastructure seemed to have reared its head: one of us remembered how in a recently completed research project the research team convened to decide on the publication outlet for the project’s final book.³ In that meeting, the national classification of publication channels assumed a central though far from an unambiguous role: publishing with a highly ranked publisher was seen as important for project prestige, a means of building reputation for and securing the uncertain futures of post-doctoral researchers and doctoral students, all linked further to the faculty’s funding distribution model of the time (that allocated additional research group funding on the basis of publication numbers weighted by the publication’s ranking in the national classification system). Although the faculty has adjusted its funding model regularly, and there was no guarantee that publishing in highly ranked series would actually bring any of the desired results, it seemed impossible to simply ignore the possibility that classifications and rankings matter.

Another episode then came onto the list, perhaps spurred on by the future-oriented, speculative nature of data infrastructures and the various rationalities that seem to sustain them – that in that vignette first emerged in the guise of uncertainties of field access. The research participants’ (see footnote iii) awareness of the high PISA scores of Finland – the country where the research team was based – lent credibility to the team and its expertise during fieldwork abroad even though at first glance the local actors also suspected that researchers from a highly ranked country would judge their work. Moreover, as one team member reported in her fieldnotes (Kauko et al. 2018), some schools at the research site connected their participation in an international research project to participation in international large-scale assessments to which local and regional authorities paid particular attention:

As the deputy principal explained later, their participation in international studies counts as the so-called project activity and gives them privileges in regional rankings, and even additional funding. So that is their main motivation to take part in our study. When schools participate in TIMSS [Trends in International Mathematics and Science Study] or PISA they get the same bonus (quoted in Kauko et al. 2018).

The items on our lists are ‘disordered sketches of lines in complex indefinite assemblages, or put differently, (…) various cogs in the machinery, without assuming that a finished and complete picture could ever be achieved’ (Petersen and Millei 2015, 12). As such, lists are ongoing and can be crossed out or added to indefinitely (Phillips 2012). Our entries do not only deal with research on data infrastructures in the academia but also school education. Moreover, these entries refer to present as well as historical infrastructures of the past. This means that they cannot be neatly bundled according to an overarching logic of operation, institutional borders or contemporaneity. This is to avoid any a priori boundary or class that would delimit listing. We note, as also do Petersen and Millei (2015), that working with lists in this way affords a means to problematize how lists have been taken up in the current climate of accountability, comparison, ranking and making commensurate, demonstrating that scientific methods are not solely the province of academics and that there is a permeable boundary between scientific methods and technologies of governance.

Moreover, the listing exercise as a decidedly affective engagement alerts us to the diversity of registers in which we know and interact with the social world (Lury and Wakeford 2012): we may think of and make lists when needing to ‘absorb or refract the stressful intensities of our openings and closings (…) when we need to defer things, when we cannot defer things any longer’ (Phillips 2012, 96). As we list – assemble and rub up against the vignettes and dispatches of data infrastructures – we are driven by the desire to refuse to both hierarchize or claim completeness and to refuse to homogenize or streamline what is listed, although we also see lists as sharing their inclusion into data infrastructures – or at least this is how we intuitively thought of them as belonging under this conceptual umbrella.

We see listing here as assembling heterotopias – spaces of non-relation – that have ‘the ability to juxtapose in a single real place several emplacements that are incompatible in themselves’ (Foucault 1994b, 181). As such, they resist reduction to an ‘order’ and homogeneity. As a heterotopia, a list contains different entries, but these do not form a hierarchical structure nor a seamless whole. Moreover, they do not merely signify or indicate some other spaces but have in themselves a spatializing power (Foucault 2002). Law and Mol compare each entry on a list to drawing in a sketchbook: 'Each orders and simplifies some part of the world, in one way or another, but what is drawn is always provisional and waits for the next picture, which draws things differently' (2002, 7). Lists are also heterochronias (Foucault 1994b, 182): not unlike museums, they are enactments of synchronicity with elements that have different temporal origins and rhythms (see Hetherington 2011). Our list contains entries of past events and affective attachments from our own academic career, as well as vague expectations of events that may yet come to pass.

As heterotopias, lists both find resonance with and rub up against what Lury, Parisi, and Terranova (2012) call the ‘topological rationality’ in our culture, and the emergence of both governmental and economic cultures that destabilize former notions of distance and discontinuity through new practices of commensuration and a spatial imagination in the human sciences that rejects the Euclidean and topographic notions of neutral, empty spaces – of which data infrastructures and their academic study establish a clear case in point (Ruppert 2012; Saari 2012). We think of listing as a topological play that destabilizes the notion of a pre-given ‘order’ or a grid which delimits which entries can ‘fit’, ‘belong’, be ‘close’, or ‘coterminous’ with one another and, instead, establishes its own distances and parallels.

We find that thinking about listing as a heterotopic (and heterochronic) practice has the effect of unsettling our own relation to data infrastructures. Following Foucault (1994b) we think of heterotopic lists as analogous to mirrors. Mirror is a heterotopia par excellence ‘in the sense that it makes this place I occupy at the moment I look at myself in the glass both utterly real, connected with the entire space surrounding it, and utterly unreal – since, to be perceived, it is obliged to go by way of that virtual point which is over there’ (179). Thus, by writing down and reading a list of our own experiences, we can see ourselves becoming inscribed, in a fractured way, in a ‘place’ where we actually are not, implicated in a disordered list of entries about infrastructures that lack a clear onto-epistemological logic of categorization. As such, our own ‘experience’ becomes less present, less interior and personal, whereby, the place we look at reflects an air of unreality, a non-presence, Saari (2012). The list/mirror as a space of representation thus renders unstable the relations between the ‘here’ and ‘there’, of interiority and exteriority, presence and absence. A list can at once pin us down and displace us (756).

The listing exercise not only shed light on the provisional and sketchy, and also somewhat connect(ed/ing) and generative nature of data infrastructures, but also alerted to how the possible connections are constructed by the researchers and the researched alike. Thus, they do not exist a priori but only emerge at sites of situated practices (Amin 2002). Moreover, the ‘insides and outsides are continuous, where borders of inclusion and exclusion do not coincide with the edges of a demarcated territory, and where it is the mutable quality of relations that determine distance and proximity, rather than a singular and absolute measure’ (Harvey 2012, 78). This approach helped us to take seriously the relational and topological nature of data infrastructures and echoed the notion that infrastructures are distributions: they are and perform distributed activities along social, technical and institutional axes and over space and time (Bowker et al. 2010). In the next section, we move on to explore two identifiable dispositions of data infrastructures that we have analysed on the basis of our list items and mindful of the extensive body of literature on the emergence and proliferation of data infrastructures.

Infrastructures as assemblages

Provisional and dispersed

As we explained earlier, our research on data infrastructures has led us to conceptualize infrastructures as assemblages (echoing researchers such as Gulson and Sellar 2019; Hartong 2018, Kitchin 2014, among others). Here we point to the Deleuzo-Guattarian use of the concept as it highlights the performativity and constant coming-together of symbolic, material and social elements in data infrastructures.⁴ Thus, there is no ‘essence’ within or fixed borders of the fringes of infrastructures (Sellar 2015a, 4–5). Picking out an entry from our list, a study on the establishment of a Finnish evaluation system in the 1970s, Saari (2012) sought, following an ANT rationale (Latour 2005, 29–36), to trace carefully the ways in which actors form, reproduce and mobilize infrastructures as networks. This commenced by charting how official documents position different institutions and their interrelations in evaluating the preparation and implementation of compulsory school reform – the National Board of Education and its evaluation office as central coordinator, national research institute of education, schools, databanks and so on. These institutes would produce, gather, circulate and analyse data on school reform experiments and achievement tests. This infrastructure formed a system of visibility which delimited what could be seen, thought, contemplated and acted on as existing and relevant phenomena of primary education. Evaluation data should focus on observable and measurable behaviour that could be compared to behavioural objectives. The evaluation infrastructure emerged as a feedback system consisting of four stages: planning, instruction, results and evaluation. This process formed a closed circuit of information, where evaluation was linked with feedback on the results and new stages of planning. In order for this infrastructure to operate, the production of data had to be standardized, which limited forms of behaviour – from researchers and administrators to teachers and pupils. Test instructions had to be administered in a standardized way, likewise the completion of test papers, which then had to be stored together and subsequently analysed according to strict statistical measures, gathered into statistics that formed a plane of comparison (on standardization and objectivity-making see also Piattoeva and Saari 2018; Williamson and Piattoeva 2019).

This example substantiates the insight by Gulson and Sellar (2019) that

infrastructure is not simply an underlying arrangement of technical objects and systems, but also includes a variety of more intangible elements and practices: habits of thought, subjectivities, social practices and so on. Infrastructure is thus constituted from, and constitutes, social relations, cultures, desires and beliefs, and in relation to governance, it is constituted by, and constitutes, various modes of both centralised and dispersed power. (Gulson and Sellar 2019, 3; see also Kitchin 2014.)

The amalgam of material configurations, standardized practices and forms of subjectivity form a dynamic data economy: just as money does not simply have intrinsic value, data only come to signify something and become relevant once mobilized – compared with other data and circulated between different actors. Ideally, such an economy should be annexed to ever new institutions and their infrastructures in an incessant movement of ‘deterritorialisation’ and ‘reterritorialization’ (Deleuze and Guattari 2008, 559–562): evaluation infrastructure disconnects existing practices of documentation and assessment from common institutional spaces and merges them with an ever-expanding standardized network of evaluation practices. Achievement test data would be connected to international testing networks such as the IEA, and national databanks could provide important information on pupils for purposes of teacher education and in-service training and as a basis for discussing policies and legislating in government and education and labour ministries. (Saari 2012.) This constant forging of new links and new connections rendered the limits of infrastructures elusive. Tracing them descended into an incessant ‘stuttering’ (Deleuze 1998): infrastructures plug and bifurcate into this, and this, and this … These can be partially controlled from a centre of calculation, but often entail connections and effects that are not designed in advance and are responsive to institutional, cultural, and political dynamics in different scales. For instance, in their study on law school ranking system in the United States, Espeland and Sauder (2016) noticed the USN (U.S. News and World Report) infrastructure of ranking being composed of other tests and data infrastructures. Moreover, law school ranking scores became connected to an infrastructure determining faculty funding. Yet the USN rankings also promulgated effects into unexpected areas, for instance, affecting the probability of a law school researcher getting published in a prestigious journal (116–117).

But even as infrastructures can be seen as dynamic amalgams dovetailing into each other – factually or fictively – technically, discursively, affectively and in other ways, this also entails that the functioning or malfunctioning of one infrastructure, or some of its constitutive elements, necessitates and affects the functioning – or malfunctioning – of another. For instance, Piattoeva (2016) shows how the perceived mal/functioning of the infrastructure of the national school graduation examination that powers the databank on school performance data resulted in authorities introducing a system of video surveillance to steer the examinations, giving rise to ‘data on data collection’. As this example connotes, the expansion of data infrastructures, especially when related to the prerogatives of performance measurement and accountability, invokes resistance and subversion (see also Gorur 2018). The irony is that such re-actions, while pointing to the unstable and relational nature of data infrastructures, add to the assemblage and make it more intrusive, implying that assemblages are processes of making and unmaking (cf. Jackson and Mazzei 2013) and their expansion and durability are contingent upon failure.

In addition to their ability to ingest malfunctions, the shifting boundaries between the inside and outside emanate from and engender the multiplication of relationships. Harvey, Jensen, and Morita (2017) defined infrastructures as ‘extended material assemblages that generate effects and structure social relations, either through engineered (i.e. planned and purposefully crafted) or non-engineered (i.e. unplanned and emergent) activities’. Seen thus, infrastructures appear doubly relational as a result of their simultaneous internal multiplicity and their connective capacities outwards (Harvey 2017, as cited in Harvey, Jensen, and Morita 2017). Michael Power’s (2004) differentiation between first- and second-order measurements constituting a connection between infrastructures helps to shed light on just one aspect of the amplification of relationships generated by the multiplicity of ‘internal’ and ‘external’ capacities. First-order measurements in education would be those everyday quantifying practices of grading and the more systematic achievement testing procedures at schools. The second-order measurements constitute operations performed on the first-order data, whereupon test results are aggregated, cross-tabulated with other data (population, socioeconomic indicators) and related to governmental demands and aims of equality, effectiveness, reform, etc. This requires new kinds of specialists to ensure the quality, storage and communication of data and to explicate their significance in relation to governing the education system. In this, the focus on the elusive and complex relation between the phenomena and the simplified nature of numbers is easily obscured and obfuscated (cf. 773).

Second-order measurements point to the potentially broad circulation and use of data in governance – to their elasticity and combinability, and their applicability in new contexts and for an expanding range of governmental tasks (Piattoeva 2015). And as Radhika Gorur (2016) discusses, assessment data also circulate widely in academic circles: there is a growing body of academic work feeding on the secondary analyses of international large-scale assessments. Secondary analyses enabled by data sharing through digital databases rely on data being flat, mobile and promiscuous, that is, ‘traveling across times and spaces speedily and combining freely and without restraint with other similarly displaced objects, to produce knowledge that may be mathematically defensible but perhaps ontologically absurd’ (651). As she explains further, ‘databases made up of ontologically impoverished objects provide a kind of surface infrastructure (as opposed to a strong foundation) on which researchers can skate with speed and efficiency, and create apparently solid science through defensible calculations’ (665). However, this is not simply a process in which reality is reduced, but these actions, instead, leave us with more: both ‘complex qualities subject to commensuration’ and ‘simplified representations produced through this process’ are added to the world (Sellar 2015b, 132).

Technical improvements and academic engagements – both debunking and affirmative – add to reality confirming that the focus of critique needs to shift from the technical or political aspects to the co-constitutive and generative capacities of data infrastructures (as Gorur 2017 has also called for). We are reminded here of Bruno Latour (2004), who directs us to the forms of critique that add to reality by re-assembling the many participants that gather to make a thing exist and endure. If we take the direction of critique proposed by Latour (2004) towards – not away from – the gathering, we will at some point see how infrastructures summon a bewildering number and variety of ‘matters of concern’, that is, embedded gatherings of heterogeneous political, material and affective interests, objects, forces and connections.

These are definitions that we have found to resonate with our research on data infrastructures. Sticking with the notion of data infrastructures as assemblages, we move to the next entries on our list, which bring the analysis of infrastructures closer to the entanglement between socio-material and affective determinants, and this also involves examining the researcher and the researched as being affectively inscribed in the infrastructures.

Affective attachments

The move towards digital monitoring and quantification has significant implications for shaping academic work practices and identities. As part of this use of data analytics, academics are encouraged – and indeed, in many cases compelled – to collect data about their research and teaching practices, reflect on the apparent insights these data afford them and work to make changes so that their data can be improved. The academic quantified self has become a key feature of contemporary higher education (Lupton, Mewburn, and Thomson 2018). Academic workers are subjected to a plethora of measurements and metrics: from Google Scholar listings of citations for each of their publications to student evaluations of their teaching to national research evaluation exercises, all of which serve to continually formulate ‘metric assemblages’ of different kinds (Burrows 2012).

These examples serve to testify that infrastructures are woven into our daily practices (Bowker et al. 2010) as academics: they operate on and through the body and resonate in the reflexive relations of self to self. Espeland and Sauder (2007) approach the constitutive nature of data through the concept of reactivity: ‘Because people are reflexive beings who continually monitor and interpret the world and adjust their actions accordingly, measures are re-active. Measures elicit responses from people who intervene in the objects they measure’ (2007, 1). While we agree to this definition, we also think that re-activity as a rational activity is not able to capture the full effect of data infrastructures. The focus on affects as vital, visceral forces helps to broaden the view of the subjective resonances of infrastructures from the merely rational and reflexive (Sellar 2015b; Staunæs and Pors 2015; Brøgger and Staunæs2016).

It may sound counterintuitive to talk of affects in the same sentence as infrastructures. At least one might think that they repel or suppress emotional reactions. Bureaucratic Grey is the colour of infrastructures – as described in Star’s (1999) sarcastic lamentation of infrastructures as ‘boring’ and ‘unexciting’ for data engineers, civil servants, policy-makers and students as well as the scholars implicated in them. Yet we propose that taking insights from affect studies may open up novel vistas for the study of infrastructures (see also Brøgger and Staunæs 2016; Sellar 2015b).

The recent ‘affective turn’ in political studies (see, e.g. Hoggett and Thompson 2012; Wetherell 2012) has pointed towards the political weight of affects. This challenges the way affects and emotions are territorialized as residing ‘inside’ the individual, as standalone entities amenable to introspection and expression (Ahmed 2004a, 8–9). Ahmed claims that they are not simply ‘within’ or ‘without’ but ‘(…) create the very effect of the surfaces or boundaries of bodies and worlds’ (Ahmed 2004b, 117). Thus, there is nothing ‘natural’ or ‘given’ in affects. Ahmed therefore asks us to bypass questions about what affects are, to ask what they do, that is, how they are used, circulated, differentiated and stabilized and how they attach individuals to communities and spaces (Ahmed 2004a, 4, 2004b, 119). They also form an ‘economy’ (119) in which they may be mobilized for political and economic purposes. Indeed, affects operate much like capital, the value of which need have no inherent positive value, but as capital, ‘it is produced only as an effect of its circulation.’ (120).

Data infrastructures are co-constitutive of affects: infrastructures produce, exhilarate, suppress, circulate and disperse them, yet they may also be the fuel indispensable to sustaining the very existence of infrastructures. Indeed, as Sellar (2015b) points out, ‘the use of (…) data for governance purposes depends on emotional or felt effects that data and associated judgments have on those whose practices are made commensurate in order to be compared and evaluated, sanctioned of rewarded’. Brøgger and Staunæs (2016) take the example of shame as an emotion that is formed, sustained and circulated through standards. Analysing a project that seeks to render national higher education systems across Europe comparable in relation to the Bologna process, they identify an implicit naming-and-shaming effect indispensable to standardization. Naming here refers to the way a scorecard system produces a plane of comparison which makes it possible to identify countries’ achievement in comparison to others through colour coding (from dark green = excellent performance, to red = little progress has been made). The participants in the comparison have an affective investment in the infrastructure of comparison and thus it is also an indirect shaming system that incites countries with a red code to consciously implement given standards so as to be relieved of the disgraceful colour tag (230).

Another example can be read from Espeland and Sauder (2016) aforementioned study on law school rankings. While they do not explicitly focus and theorize on affective investments, they report how ranking systems generate and regenerate affective responses, including constant fear among the deans of a drop in rankings. Deans also have to manage the affects that circulate within the constituency prior to and after the publication of rankings. For instance, a penetrating ‘demoralization’ was witnessed in the whole law school staff after an unexpected drop in the table (127–129). We notice similarities in the way our own publications, as they become annexed to infrastructures like Academia.edu or Researchgate, also convey and circulate affects. These infrastructures form planes of comparison in which one’s own subjectivity – one’s ‘worth’ as a researcher compared to others – becomes inscribed. Listings of publications become readable as signifiers of success-or-failure, of a ‘research profile’ and dynamism, thereby inciting affects that move us to seek recognition and to avoid a feeling of inadequacy due to being excluded from a network. Here, the economy of data becomes amply imbued with the economy of affects – that are constantly mobilized and aligned with each other: the site, with its notifications of new publications, new readers and contacts, may impart a feeling of having to constantly feed the system with new publications in order to be acknowledged as an active researcher. Here, the impending threat of inadequacy is spread across subjects, forming negative attachments to standards and the infrastructures that make them effective.

Yet infrastructures do not merely feed on negative attachments. Feelings of inadequacy may alternate with a sense of pride at achieving recognition and feedback from colleagues, making new acquaintances (see also Ball 2003, 221). Indeed, Academia.edu actually feeds on and uses these affects, constantly emailing us notifications such as: ‘Someone just searched for you on Google. To see what city they came from and what paper they viewed, follow the link below’. To see what texts cite our articles, and to ‘track your growing reputation’, we are urged to upgrade our subscriptions by paying an extra fee. Thus, curiosity and ambition are the affective-emotional fuel that generates data economy and keeps data on the move.

The above case makes it clear that as academics, we are not inscribed to be mere passive prisoners of the system, but assumed to be autonomous agents in programmes and strategies of ‘soft governance’ (Lawn 2006). Affects form an important relay between programmes and strategies of government and autonomous individuals: ‘to govern through how someone is sensing oneself and to manipulate how they relate to this sensing. Not in a predictable, controlled way, but by affecting in a dynamic, agenda-setting manner.’ (Brøgger and Staunæs 2016, 230). These examples show how affects contribute to ‘sticking’ subjects, standards and infrastructures together, forming a collective, the members of which enjoy a degree of autonomy.

Yet what is rarely acknowledged is that academics stand at the intersection of different data infrastructures, each standardizing academic practices, circulating data and affects through slightly different rationales. For instance, the Finnish national journal classification system JUFO classifies journals and publishers according to a tripartite system: 1: basic; 2: leading; 3: top (Tieteellisten seurojen valtuuskunta 2019). An education journal tagged with a 1 may be at odds with the esteem it is accorded in certain circles, and it may also have a high impact factor. In selecting a possible outlet for our own publications we are torn in multiple different directions, which may cause anxiety as to which logic should be followed and how our resolutions may affect our own career.

Anxiety is often deemed the feeling most characteristic of the neoliberal age (Salecl 2004; Krce-Ivančić 2018). It is situated on the threshold between an inchoate, unconscious affect and a conscious emotion. What is characteristic of anxiety is that it does not have a specific object: it is a hovering feeling imbued with a sense of foreboding. Something is going to happen, but it remains unclear just what that will be. Anxiety is caused by the constant altering and incessant adjustments of what is made visible, countable and legible. The perpetual shifting of data infrastructures to make explicit and instil anticipation, uncertainty or hyper-vigilance (see Harvey, Reeves, and Ruppert 2013) powers the angst side of infrastructures. As in the vignette referring to the project meetings where publication rankings seemed indispensable, data infrastructures may produce subjectivities through ‘anticipation of neoliberalism’ defined by Molé as ‘a psychological process at the nexus of expectation, affect, and temporal imaginings’ (2010, 47). Krce-Ivančić (2018) argues that the neoliberal subject is assumed to be autonomous and self-reliant, to plan ahead, yet at the same time, he or she is caught in the middle of inconsistent and ever-changing demands and an ominous sense that his or her own fate remains utterly contingent, as there is no instance that is in control or responsible for the system as a whole. It is therefore typical, as our own mundane encounters with data infrastructures serve to testify, that a sense of achievement of getting published and achieving citations alternates with the anxiety of never being sure about which rationale to follow.

Moreover, anxiety may be caused by conflicting demands between the injunction to follow the rationales of infrastructures and a more traditional academic ethos of doing one’s work out of a sense of vocation and curiosity or out of a wish to serve society. Gregory Bateson’s (1978, 206–216) term of double bind is useful here. In an interpersonal relation involving affective attachments, double binds are contradictory messages or demands – such as ‘Be yourself!’ – in which the other message, issued at a higher, more abstract level, is at odds with or negates the first. One message may be explicit, the other tacit. Double binds are not only verbal; different practices, spaces and affects may come together in, e.g., organizations to form assemblages that produce double binds that encourage autonomy and responsibility in planning one’s own work, on the one hand, while on the other introducing practices and spaces of meticulous measurement and control of behaviour (Hawes 2004).

These conflicting demands and the affects they engender and feed on also have a temporal aspect. Sellar (2015b) uses Webb and Gulson (2012) term policy prolepsis to conceptualize how education policies may create expectations of possible futures, which in turn resonate in affective registers which may contribute to realizing or diverting such futures. In the case of publication rankings, we feel caught at the intersection of conflicting incitements to plan ahead in terms of their possible rewards and forms of recognition. Whereas contradictory messages are hardly calculated effects (as mentioned above, no one is in charge of the whole), they have affective resonances functioning as ‘vectors of control’ (Sellar 2015b, 135) as they generate forms of emotionally loaded anticipation.

These experiences of anxiety, curiosity, or fleeting senses of pride and achievement often escape the existing analyses of the processes of commensuration in and between infrastructures. These interstitial affects emerge on the fringes or at unexpected points of contact between data infrastructures. They cannot be located within the dimensions of subjective-objective, figure-ground in academic research. Although transmitted through bodies and signs, binding together an assemblage of symbols, materialities and subjects through positive and/or negative attachments, they never blend different parts into a cohesive whole.

Shore and Wright (2015, 22) have emphasized the ‘alarming easiness with which organizations and individuals have adapted to the calculative, performative rationality despite exhibiting critical views of crude measurements’. Performance measurement is said to work on and (re)-constitute the subjectivity of (academic) actors while their proliferation ironically rests on our own complicity and agency. A glimpse at the economy of affects that pervade data infrastructures may shed light on how performativity is imbued with affective investments. They do not necessarily transform us into submissive academics willingly complicit in the performative aspects of academic work. Instead, infrastructures may operate despite, and sometimes, because of negative attachments as well as of positive ones.

Conclusion

There is currently a host of influential texts exhorting us to view infrastructures with a decidedly critical eye. This is not despite, but because of their seemingly innocuous guise. Mukerji (2010) and Patrick Joyce (2003) both claim that material structures like infrastructures easily ‘disappear from consciousness when they are accepted and taken for granted, creating an illusion of freedom …’ (Mukerji 2010, 403). Hence, infrastructures need demystifying intervention by critical research. Kitchin and Lauriault (2014) note that the normalizing effects of infrastructures need to be constantly and recursively reaffirmed through ‘implementation, management and system governance’ (23). In the same vein, Susan Star (1999) called for constant attention to be paid to the ‘forgotten, the background, the frozen in place’ and for a constant search for understudied, unconceptualized areas in infrastructures. The aim is to unfreeze taken-for-granted master narratives underlying infrastructures, to bring to the surface what is invisible, or to unpack black boxes along with the paradoxical, seemingly irrational aspects of building, using, repairing and expanding infrastructures (384–387).

While we wholeheartedly agree on the need to focus on infrastructures as still understudied conduits of power, we take the view that such research may benefit from elaborating on the notion of the impossibility of exteriority as a leitmotif of research. In particular, we find it useful to approach data infrastructures as ‘matters of concern’ (Latour 2004). This approach does not so much seek to uncover and demystify the mechanisms undergirding the power effects of data infrastructures as to add to their reality in how they are embedded in the very practice of the academic work that discusses them.

As noted, infrastructures are heterogeneous assemblages of semantic, material, temporal, affective and other elements, forming ‘a logic of unholy mixtures’ (Lecercle 2002, 54, cited in Maclure 2013, 660). What is important for us in this definition is that infrastructures are profoundly relational and their relations and effects extend exponentially, alerting to their ‘extensive temporal and spatial reach’ and the consequent ‘complexities and complications attending their open-ended relational capacities’ (Harvey, Jensen, and Morita 2017, 5). Infrastructures do not merely enable new relationalities and topologies but are also intrinsically reliant on such continuities (and ruptures). In a recursive movement, the workings of the infrastructures loop back upon societies or people, re-shaping them in turn (12).

Our approach to the listing methodology in this article has sought to divert and play with the topological thrust inherent in infrastructures and explored what is possible to think when the notion of exteriority underlying critical study is destabilized. Listing can be used as a strategy to open up new avenues of inquiry and forms of data to studying infrastructures that help to examine their decentred and evolving nature . Infrastructures are polycentric in the sense that there is no single unequivocal centre that controls, nor any absolute periphery or a space of exteriority (based on Star and Ruhleder 1996). Responding to the call for a ‘messier’ social science, listing does not, however, mean highlighting the disorderly nature of infrastructures. Instead, ‘(s)eeing data infrastructure as assemblage encourages us to examine the real relations that give a degree of consistency and coherence to functional arrangements of disparate things …’ (Sellar 2015a, 769).

The study was initiated by an inchoate feeling of walking a Moebius strip: first approaching infrastructures from the ‘outside’ as a separate set of research data, but then finding ourselves suddenly ‘inside’, infrastructures being indelibly etched in our own work as academics. As such, lists enable us to study infrastructures without the need to define their exact contours or isolate them as ‘outside’ background. Moreover, listing can be used as a diagnostic tool to identify factors that contribute to the methodological/epistemic fiction of being ‘outside’ as researchers. One such factor is the material side of infrastructures. As Mukerji notes, material structures of logistical power tend to assume the guise of apolitical, mute, neutral structures. This may also explain why they are not easily taken into consideration as affecting the conditions of possibility and forms of research. Koro-Ljungberg and MacLure (2013) also noted the binary of ‘dumb matter’ and linguistic-cultural systems that govern our understanding of ‘data’. Another factor is the scission between rational-emotional and the image of infrastructures as rational and devoid of emotional attachments. Yet, as we show, data infrastructures are intertwined with and co-constitute the affective economy of data infrastructures and these affects also play a role in how we do research. By our call to reflexivity we encouraged ourselves and colleagues to unsettle our common habits of mind that may treat data infrastructures as ‘out there’ and as an ‘it’. We concur that prevailing ontological and theoretical understandings of data infrastructures as assemblages would benefit from experimenting with post-structural methodologies that help to demonstrate and analyse the messy, multiple and illusive nature of data infrastructures that increasingly constitute education and academe across countries through entangled material and affective dispositions.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Notes on contributors

Nelli Piattoeva

Nelli Piattoeva is Associate Professor in New Social Research programme,Tampere University, Finland. Her research principally focuses on the datafication of education policy. She is interested in national and international large-scale assessments as sources of evidence for policymaking and new technologies of governance at a distance. Her previous research explored state- and nation (re)-building in the post-Soviet societies and changes in school curricula and the perceptions of ‘good’ citizenship entangled in these processes. She is also interested in the practices of knowledge production on education and the linkage between these practices, governance and (geo)politics, for instance, the impact of Cold War on knowledge-making about education and childhood in the former socialist countries . Nelli’s primary geographical focus of research is Russia and the post-Soviet space.

Antti Saari

Antti Saari is Assistant Professor in the Faculty of Education and Culture, Tampere University, Finland. His research interests include history and philosophy of education and curriculum studies. His publications address the interfaces between expert knowledges and educational policies, including how psychological and sociological discourses are translated into practices of evaluation, classroom management and the use of instructional technology.

Notes

1. Here we refer to Keller Easterling’s (2014) understanding of disposition as relative and relational tendencies, propensities or properties that are interacting with other factors. It is in other words a way to conceptualize the fluid active potentials or forms of agency and temperament of an infrastructure. Disposition is located in an object, but it realizes in interaction with other things. Moreover, if we think of disposition in terms of politics and power, it is helpful to imagine it as palpable but indeterminant repertoire ‘that refuse to cohere around a fixed position, or state their own name’ (http://kellereasterling.com/articles/disposition).

2. In this list animals are categorized as: ‘(a) belonging to the Emperor, (b) embalmed, (c) tame, (d) sucking pigs, (e) sirens, (f) fabulous, (g) stray dogs, (h) included in the present classification, (i) frenzied, (j) innumerable, (k) drawn with a very fine camelhair brush, (l) et cetera, (m) having just broken the water pitcher, (n) that from a long way off look like flies’.

3. This paper refers to the project entitled ‘Transnational Dynamics of Quality Assurance and Evaluation Politics of Basic Education in Brazil, China and Russia’ (2014–17) supported by the Academy of Finland under grants 274218, 307310, 273874. It studied national policies of accountability and their enactment and impact on local level governance and practices in schools in China, Brazil and Russia (see Kauko, Rinne, and Takala 2018).

4. The similarities between assemblage thinking and ANT include a relational view of the world, their emphasis on emergence and entanglement between the social and the material, and their topological view of space (see Müller and Schurr 2016). Despite these similarities, researchers debate whether the two approaches are actually compatible, and opinions vary . The differences have been attributed to, among others, the apparent lack of the discussion of affect in ANT. In this article, we echo Müller and Schurr (2016) call to see the two approaches as having much to gain from each other, including, as we also discuss in this article, the role of affects as important and understudied elements of data infrastructure that bind elements together. Moreover, our discussion on assemblage bridges the more concrete of the socio-material (as in ANT) with the more fluid, ephemeral and virtual qualities of assemblages as discussed in assemblage thinking. Bridging the actual and the ephemeral helped us to think about and problematize the notion of exteriority as central to this article. We thus opted for making a careful use of both resonances and apparent differences of the approaches to construct a richer analytical toolbox (see also Law 2009).