Attending to data: Exploring the use of attendance data within the datafied school

Abstract

Contemporary schooling is seen to be altering significantly in light of a combined ‘digitisation’ and ‘datafication’ of key processes. This paper examines the nature and conditions of the datafied school by exploring how a relatively prosaic and longstanding school metric (student attendance data) is being produced and used in digital form. Drawing on empirical data taken from in-depth qualitative studies in three contrasting Australian secondary schools, the paper considers ‘anticipatory’, ‘analytical’ and ‘administrative’ aspects of how digitally-mediated attendance data is produced, used and imagined by school staff. Our findings foreground a number of constraints, compromises and inconsistencies that are usually glossed-over in enthusiasms for ‘data-driven’ education. It is argued that these findings highlight the messy realities of schools’ current relationships with digital data, and the broader logics of school datafications.

Keywords

Digital data datafication schools secondary

Introduction

This paper focuses on two burgeoning aspects of education that many commentators believe are fast coming together to (re)shape the character of contemporary schooling. First, is the well-documented turn toward the quantification, and metrification of schooling – as Michael Apple (2004: 15) put it, a commitment toward “the ideology and techniques of accountability, measurement, and management”. Second, is the equally well-documented proliferation of digital technologies across school systems – often promoted enthusiastically as an innovative means of ‘enhancing’ teaching and learning, but also seen as “profoundly and quickly … reconfiguring educational governance” (Landri, 2019: 132). Together, these two developments are coalescing in the form of digitally-mediated school data – with digital systems amassing large quantities of school data that can be circulated, combined and analysed through powerful computational techniques. As Jarke and Breiter (2019: 3) describe, this ‘datafied’ mode of schooling is associated with significant potential change:

Digital data allow for the analysis of different educational practices to a degree of complexity not previously possible and to a much greater extent, as they can be very detailed, cover a more complete scope and can be flexibly combined … digital data not only serve to support decisions, but also fundamentally change the organisation of learning and teaching.

Of course, teachers, school leaders and administrators have been making use of data throughout the history of schools. However, there is understandably much support for the idea of digitally-enhanced datafied schooling amongst policymakers, industrialists, financiers, and others seeking to force the corporate reform of public education – all attracted by the promise that digital data allows us to know more about schools and schooling (see, for example, Piattoeva and Boden, 2020). Crucially, the vast amounts of digital data being generated within any school raise the possibility of having continuous ‘real-time’ information that can provide comprehensive insights into both what is going on, and what should be done next. The promise of enhanced ‘knowing’ is certainly reflected in how education data is talked about – e.g. sentient schools, smart schools, precision education, adaptive teaching, personalised learning, and so on. Such rhetoric infers an accuracy, efficiency and flexibility that are not popularly associated with mass schooling. Who would not want such qualities to be infused throughout contemporary education?

There is a growing critical academic literature that is challenging such enthusiasms. For example, a broad literature has grown up around the significance now attached to standardised measures of assessment and performance (Hardy, 2021; Lingard, 2010). The present paper adds to this recent turn by examining the changing nature of student attendance data within digitally-rich school settings. Our choice of attendance data is deliberate. Whereas most critical discussions of digital school datafication focus on the ‘trace’ data arising from online platforms (Hillman et al., 2020) or digital data generation such as behaviour tracking, emotional and affective monitoring (see Williamson, 2017), such techniques and tools are yet to be established as core elements of everyday school practice. In contrast, the recording of attendance data is one of the most familiar and deeply-embedded school data-points that there is – a mainstay technology in the rise of compulsory mass education over the past 100 years (Gleeson, 1992), and one of the first aspects of school data to be computerised in the form of ‘electronic registers’ and online roll-calls.

Indeed, school-teachers have taken class registers and roll-calls for centuries. Now, however, these prosaic data-points are entwined with the datafied logics just described. As a means of exploring the nature and conditions of how schools are now making digitally-mediated use of data, this paper addresses a deceptively straightforward question: what does it now mean to know about ‘attendance’ in contemporary school settings? How is this utterly familiar point of recording and measurement now being utilised within the data-driven milieu of contemporary schooling? What new insights, understandings, and ways of knowing are now associated with student attendance data?

Theoretical background & rationale

We address these questions by drawing on a number of cognate theoretical approaches. Most immediately, we locate our efforts within the emerging interdisciplinary field of critical data studies (see Iliadis and Russo, 2016). Work in this vein coalesces efforts from across the humanities, social and computational sciences to: “identify how the epistemological and ontological implications of data collection and data-driven processes may (re)constitute both knowledge and subjectivity” (Hintz et al., 2018: 6). Key lines of inquiry here relate to the ontological implications of digital data – in short, questioning the ways in which data now shapes how we know the world in which we live. Accompanying these concerns is a focus on the ideological nature of this data (i.e. how data convey specific sets of values, logics, interests and agendas), and the social/political consequences of contemporary data-driven society (e.g. what lines of reasoning and argumentation it reinforces over others). An underpinning focus of interest, as Hintz et al. (2018: 6) put it, is exploring how digital data “shapes the reality it measures by staking out new terrains of objects, methods of knowing, and definitions of social life”.

Of course, these are not concerns that are unique to the topic of digital data. Indeed, critical data studies tends to borrow from previous traditions. In addition to the above, we also draw on the STS-inflected literature on ‘number studies’. This body of work offers various lines of discussion relevant to the epistemological and ontological implications of conceiving the social world in terms of numbers and statistics. The number studies tradition originates in a number of seminal writings in the 1990s detailing the ways in which statistics-driven thinking arose from the increasingly bureaucratisation of nineteenth and twentieth century society (Hacking, 1990; Porter, 1996). These precedents led to contemporary ‘trust’ in our capacity to comprehend the behaviour of individuals and groups in terms of statistical regularities. At the same time, this work also foregrounds the ways in which numbers have long come to both dictate the duties of modern institutions like the school, and act as a measure of their success (Desrosières, 1998).

Of particular relevance to this present paper is the recent emphasis within ‘number studies’ research placed on examining what numbers ‘do’ and what is being ‘done’ with numbers in contemporary social settings. Such studies foreground questions over the situated ontologies of school data – i.e. “the negotiations through which some things can be labelled as data or not” (Denis, 2016). This work also draws attention to the ways in which the ontology of data is enacted – i.e. the ways in which something might be counted as ‘data’ in one context, and not as ‘data’ in other contexts (Leonelli, 2015).

Together, then, these theoretical approaches encourage us to think beyond the idea of data as a stable, fixed entity with a distinct fixed substance and character. Instead, they push us to approach data in relational terms (Lippert and Verran, 2018). This requires thinking about what is being done with data amongst different networks of actors and interests. This includes examining the meanings that people ascribe to data at various times and different contexts. In all these ways, such approaches therefore prompt us to consider student attendance data in terms of what data does within the social context of a school, as well as simply what data is (Day et al., 2014).

Exploring the ‘institutional life’ of school attendance data

In order to pursue these lines of inquiry with regards to student attendance data, we now present findings from an ongoing three-year research project investigating the datafication of secondary schools in the Melbourne metropolitan area in the state of Victoria, Australia. This research centres around three case-study schools: a small inner-city government school (Weston High School); a large suburban Catholic school (Brookdale High School); and a medium-sized private school in the outer-suburbs (Northland College). As might be expected, generating data on students’ attendance is a core part of these schools’ daily routines. Most classes commence with teachers clicking (or swiping) through online class lists, calling out student names and registering responses. This ritual is also repeated in the ‘home-room’ and ‘tutorial’ sessions that mark the beginning of each school day. All this data is fed automatically into the schools’ management systems.

Considerable efforts are also made to collect data on late arriving students. As the most traditional of the three schools, Northland continues to rely on students signing a paper-based ‘Late Book’ located at reception. In contrast, Weston boasts a large computer-based ‘Tardy Kiosk’ where late-arriving students are supposed to register directly into the school’s learning management software. However, while costing the school annual rental fees of $7000, the kiosk has fallen into disuse after numerous technical issues (“it’s a waste of money, but now we’re stuck with it”¹). Late students now report to Weston’s central reception desk, or else sneak past directly into their classes. Brookdale also operates an iPad-operated ‘Arrival Departure Terminal’ linked to the school’s learning management system (LMS). This allows students to declare their late arrival directly into the school system by swiping their ID card and inputting a multiple-choice response about the nature of their delay. Nevertheless, student interactions with this device usually involve the assistance of office administrators who staff the front-desk. Despite the high-tech technologies on show, most acts of attendance registration in all three schools remain essentially low-tech affairs.

This paper explores how the attendance data generated by such practices is subsequently used - what might be termed the ‘institutional life’ of student attendance data within these three schools. To do this, we draw on research data generated during our first phase of research in the three schools. This initial period of immersive fieldwork in Weston, Brookdale and Northland involved over 60 site visits, in situ observations and general ‘hanging around’, field notes, documentary analysis, photographing, corridor conversations, and more formally-arranged interviews with IT staff, data specialists, school leaders, and teachers. In particular, this paper draws upon a corpus of empirical data generated from interviews, computer-based ‘walk throughs’ and informal conversations with over 50 staff across the three schools employed in leadership, technology, administrative and teaching positions. The formal interviews were in-depth and semi-structured, lasting anywhere between 30 minutes and 90 minutes.

Drawing on thematic analysis of this corpus of data, the paper now goes onto explore the different ways in which student attendance data was talked about by interviewees during our discussions of the datafication of their school processes and practices. Analysis of the empirical data was rooted both in the a priori theoretical concerns outlined earlier, and a posteriori issues arising from the interview data. In this sense, analysis took what Fereday and Muir-Cochrane (2006) describe as a ‘hybrid’ process of inductive and deductive thematic analysis that allows us to fully describe the phenomenon being investigated. This involved a number of steps. First was the deductive generation of salient preliminary codes based on the research questions and salient theoretical concepts relating to datafication (e.g. data ontologies, data ideologies, data imaginaries). We then engaged in repeated re-readings of the interview corpus, paying particular attention to passages discussing student attendance data. This led to the inductive generation of data-driven codes – i.e. issues arising from the interview data. Repeated rounds of applying these codes to the corpus of interview data then allowed us to connect codes to establish broad themes, which were then themselves clustered and reduced to a smaller number of core themes that we felt encapsulated the phenomenon of attendance data use as described by our interviewees.

Our presentation of these themes mirrors the sequence of how we encountered talk about and/or actions relating to student attendance data. Thus, the following sections start with the claims and expectations raised in our introductory conversations with school leaders and managers about how data might be used to glean insights into their schools. These expectations and imaginaries are then contrasted with the various ways in which we then found attendance data actually being used by staff within each school. Finally, we turn to our later conversations with technical and administrative staff responsible for the collection, management and storage of data within each school. Echoing our interest in the situated ontologies of digital data, each of these different groups offered distinctly different perspectives on school attendance data and the purposes that it might be put. These contrasting perspectives go some way toward unpacking the complexities and compromises inherent in the datafication of school. In particular, we now go on to consider the following forms of ‘attendance data’ at large within each school – i.e.

Anticipatory accounts (how attendance data acted as a vehicle for imagining the possibilities of data-driven schooling);

Analytical enactments (how attendance data was actually being used for analytic purposes);

Administrative concerns (the logistical and infrastructural issues inherent in the generation and storage of attendance data).

Findings

(i) Anticipatory accounts of the data-driven school

We turn first to the anticipatory qualities of attendance data – in particular, expectations of what might be potentially knowable through student attendance data that was not already known. Here, we had plenty of speculative conversations about the potential of school data – many of which were framed by ambitions amongst the schools’ leaders and managers to procure sophisticated data tools (usually dashboards and portals) that might provide “quick and scaled-up” overviews of school data.²

One widely-desired capacity of these products was allowing teachers to ‘know their students’ – as Northland’s head of senior school stressed, “to generate the whole picture of the child”.³ Student attendance was one of the few data points (alongside exam grades, assessment scores and other ‘performance’ measures) that nearly all staff agreed should be included in any new tool. As Brookdale’s Director of IT recalled after a staff consultation exercise: “the information that teachers are most interested in is attendance”.⁴ However, this was not wholly supported by all interviewees. One Learning Specialist described attendance data as a ‘low-hanging’ indicator that schools “will naturally gravitate towards”⁵ – a familiar but conservative way of appropriating data.

Of interest here, then, is the promise of what might be known through attendance data. In this sense, many school leaders and teachers were eager to take student attendance data (more accurately data relating to student non-attendance) as a proxy for less tangible underlying issues. This thinking usually followed one of two distinct logics. First was a presumed commensurability between attendance data and a range of other behavioural, psychological and social issues. During our time in the schools, attendance was framed as a sign of “social disengagement”,⁶ as a means of identifying “recalcitrant students”,⁷ “students who are on the edge”,⁶ or “trouble students”.⁸ Second, were expectations that attendance data might reveal unseen patterns and trends within each school that might otherwise “slip through the cracks”.⁹ For example, in Brookdale attendance analytics were imagined as offering ‘golden’ insights into otherwise opaque student mindsets:

Having easy access to attendance data for a student opens all sorts of windows on how that student is performing. Having easy access to late arrivals or early departures is so valuable … I was visiting another school earlier this year talking about data platforms. They were able to show student attendance. The student was at school on this day but he was regularly skipping these middle two periods, just disappearing, possibly in the library. Who knew? But there’s this pattern here of the student loathed the teacher and was voting with his feet, every period 3 and 4 on this particular day when that particular subject happened. That sort of stuff is gold especially for a year level coordinator … If they can see that then they can start to do something. Because you know how it is with teenagers … they’re not necessarily going to open up about something like that. If they can get away with not confronting that problem or just getting that teacher out of their hair that’s what they’ll do. That sort of stuff will be incredibly helpful.¹⁰

The prospect of being able to identify anomalies, and then extrapolate psychological and social explanations was understandably appealing for staff in a school like Brookdale with over 1400 students. Yet, perhaps the most extreme continuation of this logic was a discussion with Northland’s principal team which mulled over the potential benefits of continual attendance monitoring:

I'd like data to be able to be looked at by staff so that they could then make a meaningful intervention. Being able to look at something and say, right in the last week this [student] has been later and later and later to their classes. So that's an issue … they've gone from being one minute late to five minutes late to 10 minutes late. Alright, this is perhaps telling us something. Or at 2 o'clock every Thursday for the last four weeks they've gone to the bathroom. Okay, what's going on? So, you know … looking for patterns”.¹¹

As we shall go on to see, the almost continuous levels of surveillance implicit in this imagined use of attendance data stood in stark contrast to all three schools’ capacity (and willingness) to actually enact such procedures.

(ii) Analytical enactments: Statutory reporting of attendance data

Indeed, these idealised data insights belie the schools’ actual uses of attendance data. Tellingly the school leaders and managers described above were not working with data to anything like the degree of specificity that they imagined might be possible. Instead, these leaders were primarily engaged with their schools’ attendance data as a statutory point of accountability. Government schools such as Weston had their overall attendance rates reported publicly on the federal government’s ‘MySchool’ website, alongside an internal ‘Principal Data Dashboard’ that was provided by the state government to benchmark Weston’s attendance rates against comparable (‘like’) schools in the state. Elsewhere, independent schools such as Northland had reporting obligations to confirm that international students’ visa requirements were being met. As the Northland’s website declared: “International students are at risk of losing their Australian Student Visa if absences remain unexplained”. In these ways, then, the prevailing administrative focus on attendance data remained firmly on fulfilling the schools’ statutory reporting obligations.

Besides these external commitments, the task of working in more detail with attendance data fell to an assortment of individual teachers - some of whom were the schools’ designated ‘data leads’ alongside others who were pursuing self-directed data analyses. Indeed, it was technically possible for most teachers in Brookdale, Northlands and Weston to ‘drill down’ into the schools’ main data systems to enquire about the attendance history of particular students. However, in practice, this task remained the preserve of ‘Head of Year’ and ‘Year Coordinator’ roles. These teachers were usually responsible for monitoring student data for repeated absences or late arrivals. That said, any aggregated cohort-wide analysis was widely agreed to be “really difficult to make meaningful sense of”:¹²

At the moment we can quite easily look up past reports. We can type in a student's name and get a PDF of however long they've been at the school. So, that's quite useful, but it's also quite clunky because you're looking at one student at a time, and going through a full PDF.⁸

The relatively inaccessible forms that this data took (a static PDF document) highlight the ways in which promises of ‘fluid’ and seamless circulation of data can be curtailed by material and organisational circumstances. In this instance, while the PDF format ensured a consistency of reporting, it also discouraged further use within the school.

(iii) Analytic enactments: Shared summaries and spreadsheets

In contrast to these rudimentary forms of data inquiry, perhaps the most involved example of working with student attendance data was Tony – Weston’s designated ‘data lead’ teacher who was tasked with what his Principal described as “high powered analysis”.¹³ As part of this role, Tony conducted annual reviews of student and teacher attendance data. During our first year of research, he was particularly interested in interrogating the school’s ‘Late’ data – i.e. data relating to students missing specific lessons and sessions within the school day. As Tony explained, “We’re trying to improve attendance in first morning session. So I’ve been using the attendance data to give them a measure of how many kids were missing tutorial”.¹⁴

In order to achieve this, Tony had created various spreadsheets and pivot tables to display relevant data to other staff. This was partially an effort to address staff complaints that Weston’s learning management system did not display specific dates (or times) when students had been marked absent or late. Using Tony’s bespoke spreadsheets, staff could flick between tabs to see absence results for specific weeks or months, and also gauge which students had ‘accumulated latenesses’ (and therefore merited an extra ‘detention’ session). As such, staff could use Tony’s data-sheets to quickly identify students that they needed to ‘chase-up’ in person. Tellingly, this data-set was only made accessible to teachers who were in a ‘Head of Year’ position (“the general teachers don’t use this”). As Tony described:

So they will look at this [column] and run their eye down. So look at all the red ones … that will mean something to them because they know all the kids from their year level. So that stage is still being done by an actual person … so this is really only about making the data accessible, I guess …¹⁴

Here, too, broader engagement with this data was neither presumed nor encouraged – with Tony’s ‘helpful’ formatting also working to limit how the data was used, and by whom.

(iv) Analytic enactments: Identifying patterns and trends

While his colleagues remained focused on engaging with attendance data in these relatively linear ways, Tony was also interested in ‘playing around’ with school datasets to look for patterns and trends. In his opinion, this was “where the data gets interesting”. During a ‘walk-through’ of these datasets with us, he pointed out a ‘pretty bad’ trend where “on any given day 2.5% of the kids are turning up late”. Moreover, in the datasets that he was showing us this figure appeared to spike at 4.5% on Mondays. As Tony explained:

[points to spike in graph on screen] “… Oh, this is interesting, I’ve broken it down by days of the week and it turns out that for every level the worst day is the day they have their ‘active’ tutorials in the gym. So lots of kids stay away that day … [and] it goes markedly upwards as they get older”¹⁴

This modest analytical insight belied considerable levels of effort and technical work-arounds. As Tony talked us through these spreadsheets, a number of salient points emerged regarding the analytical limitations of Brookdale’s attendance data. First, was the limited accessibility of the data. Regardless of the specificities of Tony’s inquiry (i.e. ‘Late’ data for the first session of the school day), Brookdale’s LMS would only allow a complete export of all attendance data for every school session. This resulted in Tony having to download “161,000 lines” of data, which he understatedly described as “taking quite a while to process”.

In contrast to its large size, this data-set was rather less impressive in terms of its quality. For example, the data did not actually specify which students had arrived late for their first (tutorial) session of the day. As Tony explained, this was due to state government reporting requirements for students to be marked as wither having attended or not attended the first session (“this goes on the Department edict that [students] have to be declared to be ‘In’ or ‘Out’ for each session”). As such, Tony’s analysis inferred a late arrival for any students who were marked absent for the first session but marked present for the second. Of course, students might have had legitimate reason for missing their entire first lesson. Moreover, in reality it was likely that some teachers will have marked late students as present. Yet, none of these ‘outliers’ were detectible from the official records.

Tony was also aware that the relatively small numbers in each class group meant that his analyses were sensitive to ‘significant variations’ resulting from the actions of only “a small number of individuals”, or perhaps occasional efforts by individual teachers to tackle lateness. Tony also pointed to the limitations of making these analyses easily accessible to staff. He had chosen to do this on Google Sheets rather than his favoured Excel – primarily because Google Sheets could be shared on a continually-updated form amongst a large group. The trade-off for this, however, was that “it’s not as sophisticated and it’s a lot slower”. He also conceded that there was simply too many lines of data for most teachers “to deal with”.

Tony was also keen to stress the need for human oversight of any analysis of this data. In his opinion, this was relatively blunt data that required a great deal of contextualisation before it could be analysed: “there’s always human discretion to it”. In particular, Tony stressed the need for local sense-making in order to interpret any trends and patterns. For instance, he speculated that the 4.5% spikes could well indicate varying staff attitudes to enforcing the school’s lateness policies. This was a contextual variable that was not quantifiable:

It’s probably also a measure of how teachers treat [attendance]. So if you’re really casual about kids wandering in late, then they’re going to think oh well, that’s okay, I’ll wander in late when I feel like it. Whereas if you make a fuss about it and talk to them afterwards, they feel more accountable. So it’s probably a reflection of that.¹⁴

From Tony’s point of view, then, these data points lacked the contextual detail required to draw inferences from what otherwise seemed to be a striking pattern in the data. For this teacher, at least, such findings were of little practical use.

(v) Administrative concerns: Inconsistent reporting

In contrast to these analytical accounts, we now turn to how student attendance data was described in our later conversations with the staff responsible for its production and collation. These were administrative and technical staff concerned primarily with what might be termed ‘back-end’ perspectives in comparison to the analytical work previously described. In contrast to their colleagues in teaching and school management positions, these interviewees raised a number of infrastructural and logistical issues that shape the nature and form of attendance data in the schools.

First, was the extensive range of administrative inconsistencies in the recording of student attendance. As Brookdale’s office manager put it: “the quality of data depends on how the students and the teachers enter the data”.¹⁵ This was certainly evident in our informal conversations around the schools. Some teachers would admit that roll-call mistakes were inevitably made on occasion. Others would admit to exercising discretion and not always recording students as late-attenders. For instance, Weston’s official policy was to record ‘lateness to class’ in five-minute increments. However, few teachers reported using this level of precision, meaning that the data being used by Tony was reduced to a blunt binary of ‘Late’/’Not Late’.

Alongside these inconsistencies were the obfuscatory efforts of some students. This was most immediately obvious to the schools’ administrative and office staff, who dealt with recording ‘Lates’ and also legitimating absences as ‘authorised’ as opposed to ‘unauthorised’. One popular way of confirming ‘authorised’ absences was via parental text message systems, with parents responding to automated messages informing them of ‘Non-Arrival at School’, and then requiring them confirm the nature of the absence (e.g. texting back ‘HS’ denoting ‘Home Sickness’). Brookdale’s office staff were aware that a few students had “tricked the system” by substituting their own mobile numbers in lieu of their parents’, and thereby “marking themselves off automatically”.¹⁵ Brookdale’s office staff were occasionally alerted to these false numbers if they had to contact the student’s ‘home’ for other reasons: “there’s been so many times where we have picked-up and thought, ‘Oh that’s not the parent - that’s the kid, or that’s their girlfriend’ . you can catch them out”.¹⁵

Unlike Weston and Northland, students in Brookdale were required to enter their late attendance at the LMS-connected ‘Arrival Departure Terminal’ in the school lobby. However, administrative staff were also well-aware that occasionally “the student just signs in and just doesn’t go to the class, he just heads off to the library or something else. So how do we track those students?”.¹⁵ Conversely, students might proceed directly to class but find that their teacher forgets to officially register them as attending the class. As one of Brookdale’s data specialists put it, “there is always a missing gap because of the human element”.¹⁶

(vi) Administrative concerns: Infrastructural breakdowns

Finally, our latter conversations with non-teaching staff also highlighted inconsistencies in the ways in which attendance data was collated and stored within the schools’ data systems. Interestingly, technical staff in all three schools categorised attendance data as primarily ‘administrative’ rather than ‘curriculum’ (these being the broad delineations that many school IT systems are loosely organised). This meant that the collation of attendance data was overseen wholly by the schools’ IT teams. In contrast, far greater care was taken with other data that was deemed to be academic-related, and therefore requiring consultation with the teaching teams that might be looking to work with the data. As Brookdale’s data analyst put it:

Attendance [data] is one of those things that do not need teachers’ input. It’s just a case of IT doing it and we say this is the way it works. But ancillary [curriculum] data needs constant interaction with the people who deal with it and what they are looking for.¹⁶

As a result, attendance data was not deemed by IT staff to require the same levels of attention (and care) as curriculum-related data that would be likely used primarily by teachers rather than administrators. This meant that instabilities and breakdowns with attendance data within the schools’ data systems were less likely to be dealt with. For example, it was known that the configuration of Brookdale’s various attendance tools made the resulting attendance data especially vulnerable to dropping out when other system elements were altered: “With attendance [data], sometimes things just drop off and they stop working. Like, sometimes the IT team change something. They fix one problem and then there’s another issue”.¹⁵ Conversely, it had only recently come to light that Northland’s data systems were pre-configured to automatically hide attendance data from previous years. This was apparent only when the school’s IT team were required to retrieve data from previous years, and found themselves mired in the rigmarole of “rolling the entire LMS back a year to get that data”:¹⁷

The LMS stores it, but it doesn’t display it … because [the LMS providers] figure no-one’s ever asked. When we found this out, we’re going “You’ve got 400 schools using your product and no-one’s ever asked you for last year’s data?. Are you nuts?”. So we can extract old reports, because it creates a PDF and that’s captured as a separate item within the system. The longitudinal data is in there, but no-one had said, 'Well, where is it? How do we get to it?' And so it’s in there but you have to roll your system back to last year to see it.¹⁸

Glitches such as these fostered a general sense amongst technical and data-focused staff that attempts to create a longitudinal sense of attendance data would be technically difficult. In Brookdale, this was exacerbated by their LMS automatically setting student records as either ‘Full’ or ‘Left’. As soon as the LMS was rolled-back then large numbers of current students were erroneously marked as ‘left’. This necessitated a complicated workaround involving the IT team running “a spare system” to allow these students’ previous data to be accessed.⁴ While this appeared to work, Northland technicians still reckoned that the system “would struggle to give you anything historical … the historical data doesn’t play well … it hasn’t been easily retrievable”.¹⁷

Discussion

Students coming to school and/or attending scheduled classes might appear to be a relatively straight-forward aspect of school life to record, quantify and analyse. Such data has been recorded in classrooms for centuries, and ‘attendance’ remains one of the data-points that educators immediately raise when discussing matters of data analysis and the promise of the data-driven school. Nevertheless, focusing our attention on this mundane form of datafication has foregrounded a number of constraints, compromises and inconsistencies that are usually glossed-over in enthusiasms for the ‘data-driven’ school. These findings can therefore tell us much about schools’ relationships with data, and the broader logics of school datafication in general.

Indeed, this paper points to a number of constraints and compromises inherent in even this seemingly basic instance of datafication. First, then, is the remarkably limited nature of these schools’ data cultures and data capacities. These schools were not home to the generation of comprehensive data-sets, or any subsequent form of ‘big data’ analysis. As such, student attendance data continued to be afforded a privileged position within the schools’ data practices “because of their capacity to be enumerated and the dominant status afforded to these numeric representations” (Hardy and Lewis, 2018: 673). Despite the imagined possible uses of this data, there were few other data-sets that the schools’ attendance data could be correlated with, or modelled against. As such, attendance data was largely being used in the ways in which always had – i.e. for identification of individual student transgression. The ‘analytics’ that were actually being deployed here were simple frequency counts, colour-coding and modest cross-tabulations.

Second, however, this is not to imply a particular lack of data expertise on the part of individual staff. As it transpired, it would not have been technically or statistically possible to do much more with the forms of attendance data being generated in each school. These were relatively small, compromised data-sets that lacked the trustworthiness or granularity to underpin any kind of rigorous large-school modelling or prediction. Some data were inevitably missing, mis-entered or mis-reported. Our findings therefore mirror Star and Lampland’s (2009: 14) point that all data-sets are incomplete and constructed through compromises – in other words, an unavoidable aspect of formalization or standardization is its “always already incomplete and inadequate … character”.

Third, was the largely inadequate and unhelpful forms of digital technology that were failing to ‘support’ the generation and use of data within the school. These took a number of forms – from dysfunctional attendance kiosks through to the relative ease with which students could circumvent SMS reporting procedures. One key point was the limited capacity of schools’ data systems to collate, store and share data with staff. This was not data that was ‘readily available’ for use (Loukissas, 2019). Schools’ systems were not configured in ways that presumed staff might want to access data. Outputs took the form of static, non-editable and therefore ‘closed’ PDF reports (see Wieczorkowski, 2019). Conversely, even these inadequate forms of data systems required substantial amounts of work to maintain, sustain and repair. As Lippert and Verran (2018) observe, the compensatory work that takes place around digital data to ‘keep the show on the road’ is usually glossed over in discussions of datafied society – yet “human actors, and potentially artificial actors, too, are partially well aware of tensions and frictions within their numbers, data or algorithms” (p. 9).

Above all, was an over-riding sense that people working within these schools (in contrast to any imagined possibilities) retained relatively limited ambitions for their actual use of data. As the paper showed, most school staff remained relatively fixed in what they wanted to do with the student attendance data. Tellingly, any broader insights from these analyses were yet to translate into follow-up action within the school. Despite the consistency of his data analysis, Tony admitted in subsequent conversations that the data insight of students not attending Monday morning classes had not been followed-up by school leaders. The primary institutional imperative for this data was to support the immediate task of identifying individual students and dealing with them on a case-by-case basis. There was not enough time to pause and consider any ‘bigger picture’ revelations.

Finally, returning to the theoretical conceits of this paper, our findings do shed some interesting ontological light on the nature of school digital data. Above all, we find student attendance data to be highly social in nature - a result of social organisation as well as technical processes (Dourish and Mazmanian 2012). All the forms of attendance data detailed in this paper clearly result from the interactions between people, places and technologies. Any piece of ‘data’ therefore derives from networks of people, technologies and physical objects within schools, all of which are entangled in social practices, interactions, language and other forms of representation. Conversely, we have also seen how attendance data remains a fulcrum of outside-school relations between the school, family, government, law and education marketplace (see Gleeson, 1992) - thus reflecting “complex multi-scalar relationships with numbers at a national, regional and local levels” (Hardy, 2015: 36). This raises a number of important factors and issues which point to tensions between the overall institutional logics implicit in this data, and the local enactments that produce and sustain the data in schools.

For example, our findings highlight the ideological nature of schools’ uses of student attendance data – i.e. the ways in which the data were being gathered and used conveyed specific sets of values, logics, interests and agendas. Here two different logics were apparent in our findings – both of which relate to the enforcement of institutional control over students. In one sense, schools’ data uses were wrapped around logics of care – i.e. supporting certain forms of behaviour. In another sense, schools’ data uses were imbued explicitly around logics of control – i.e. relating to data-driven surveillance and the aim of governing or managing behaviour. Crucially, it is important to see these logics in terms of institutional intentions for the individual students who make up ‘the school’. For example, both these logics of care and control are concerned ultimately with supporting forms of behaviour that the schools consider as being in students’ best interests (i.e. coming to school and attending lessons).

The political consequences of this data use, therefore, remain firmly centred on the control of students’ bodies – what can be referred to in Foucauldian terms as biopower. These schools’ uses of ‘digital’ student attendance data therefore relates to the traditional logic of measuring attendance within a school system where attendance is compulsory. In this sense, any lack of attendance is a deviant act, and therefore taken as a problem (and likely indicative of further underpinning problems). The ways in which these supposedly more flexible and perceptive forms of student attendance data are being used in schools does nothing to disrupt this logic – in fact, it perpetuates it – “these numbers are not inherently ‘objective’ but … deeply complicit in how the phenomena to which they relate are known and understood” (Hardy and Lewis, 2018: 673).

However, in contrast to these institutional logics are the various ways that our investigations pointed to the situated ontologies of school data – the negotiations and relational actions that led to data being generated in specific ways (Denis, 2016). Perhaps of most interest here was our general lack of finding that “data-responsive logics have come to dominate schooling practices” (Hardy and Lewis, 2018: 672). Instead, our findings point to all manner of compromises, incidents of non-compliance and reconstitution of data-driven logics. For example, we found some teachers not recording students as ‘late’ (for all manner of reasons), and other teachers not recording to the degree of specificity of five-minute increments. Elsewhere we found data systems being configured in the assumption that people would not want to access historical or specific sub-sets of data. These forms of school data are therefore a long way from David Beer’s (2019) description of the automated data gaze – detached, mediated and exercised at a distance by automated analytics and other data systems. Whereas Beer (2019) talks of digital data being used in ways that are “speedy, accessible, revealing, panoramic, prophetic and smart”, the ways in which our three schools approached their student attendance data was notably slow, retrospective, confirmatory and inaccessible.

Any discussion of school datafication, therefore, needs to be set against the observation that schools do not appear to be particularly motivated to respond to (or even look for) novel insights and unexpected patterns and correlations in their data. In short, school data is not a place for surprises, counterintuition and ‘outside-the-box’ thinking. On one hand, this analytic conservativeness belies one of the perceived main qualities of computerised data analysis – what David Beer (2019) describes as “a constant driving curiosity”. Nevertheless, this ambition is clearly compromised when enacted in school systems that might not be particularly ‘curious’ by nature or design. Instead, schools tend to be ordered, procedural and routinised entities that operate within systems that have very fixed notions of cause and effect. Put crudely, the prevailing focus of our three schools’ data uses was ensuring that students were attending school and then engaging in learning tasks. In this sense, there was little institutional incentive to gain insights from using this knowledge to explore further issues. Instead, school staff are interested primarily in using data to help them continue acting in broadly the same ways that they are accustomed.

Conclusions

This brings us to the central conclusion of the paper – in short, the lack of radical difference or change that is associated with the ways in which our schools were actually making use of the masses of digital data now in existence within all schools. In this paper we deliberately focused on attendance data as one of the most frequently collected and scrutinized ‘data-points’ in contemporary schooling. The fact that this common form of school data was being used in the limited ways reported, can therefore be seen as raising a number of key issues applicable to school data in general. While others have talked of the “ambivalent and at times contradictory consequences” (Jarke and Breiter, 2019: 5) arising from the datafication of education, perhaps the greatest ambivalence in our case study schools was the general lack of change or transformation. Using the example of attendance data, our findings illustrate how any potency of digital data to alter what is known in schools (and alter how that knowledge is acted upon) quickly fades into the background. The presence of digital data relating to student attendance does not lead to any notably precise decisions, or provide insight into otherwise imperceptible patterns and trends.

Yet, the fact that schools are appropriating data in ways that allow them to continue doing what they always have is not necessarily a problem or shortcoming. Indeed, as Yanni Loukissas (2019) reasons, it is important to remember that all digital data have complex attachments to place – not least in the local sources that they are generated from, the local data infrastructures and interfaces they circulate through. In this sense, while digital data often enable remote third parties to make decisions ‘objectively’ and ‘from a distance’, Loukissas (2019) argues that data must be understood in context of the people and places that lie ‘beyond data’. It seems that school data – as with any information infrastructure – implicitly faces what Star and Ruhleder (1996: 111) describe as “tension between local, customized, intimate, and flexible use on the one hand, and the need for standards and continuity on the other hand”. In this sense, it is understandable that our schools were not attempting to do anything radically different with their digital attendance data. At least for the time being, local contexts of schooling seem to be outweighing any broader promises of data-driven transformation.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Australian Research Council [DP190102286].

ORCID iD

Neil Selwyn

Notes

Descriptions of interviewees quoted directly in the paper:

References

Apple

(2004) Creating difference. Educational Policy 18(1): 12–44.

Beer

(2019) The Data Gaze. Newcastle Upon Tyne: Sage.

Day

Lury

Wakeford

(2014) Number ecologies. Distinktion: Journal of Social Theory 15(2): 123–154.

Denis J (2016) Data and its invisible work. Science + technology by other means. Paper presented to 4S/EASST Conference, August, Barcelona – https://halshs.archives-ouvertes.fr/halshs-01364311/document

Desrosières

(1998) The Politics of Large Numbers. Cambridge, MA: Harvard University Press.

Dourish

Mazmanian

(2012) Media as material. In: Carlile

Nicolini

Langley

Tsoukas

(eds) Perspectives on Process Organization Studies. Oxford: Oxford University Press.

Fereday

Muir-Cochrane

(2006) Demonstrating rigor using thematic analysis. International Journal of Qualitative Methods 5(1): 80–92.

Gleeson

(1992) School attendance and truancy. The Sociological Review 40(3): 437–490.

Hacking

(1990) The Taming of Chance. Cambridge: Cambridge University Press.

10.

Hardy

(2015) I’m just a numbers person. International Studies in Sociology of Education 25(1): 20–37.

11.

Hardy I (2021) The quandary of quantification: data, numbers and teachers’ learning. Journal of Education Policy 36(1): 44–63.

12.

Hardy

Lewis

(2018) Visibility, invisibility, and visualisation. Pedagogy, Culture & Society 26(2): 233–248.

13.

Hillman

Rensfeldt

Ivarsson

(2020) Brave new platforms. Learning, Media and Technology 45(1): 7–16.

14.

Hintz

Dencik

Wahl-Jorgensen

(2018) Digital Citizenship in a Datafied Society. Hoboken, NJ: John Wiley & Sons.

15.

Iliadis

Russo

(2016) Critical data studies. Big Data & Society 3(2): 205395171667423.

16.

Jarke

Breiter

(2019) The datafication of education. Learning, Media and Technology 44(1): 1–6.

17.

Landri

(2019) Cartographies of the digital governance of education. In: Gorur

Sellar

Steiner-Khamsi

(eds) World Yearbook of Education. London: Routledge, pp. 132–147.

18.

Leonelli

(2015) What counts as scientific data? Philosophy of Science 82(5): 810–821.

19.

Lingard

(2010) Policy borrowing, policy learning. Critical Studies in Education 51(2): 129–147.

20.

Lippert

Verran

(2018) After numbers? Science & Technology Studies 31(4): 2–12.

21.

Loukissas

(2019) All Data are Local. London: MIT Press.

22.

Piattoeva

Boden

(2020) Escaping numbers? International Studies in Sociology of Education 29(1–2): 1–18.

23.

Porter

(1996) Trust in Numbers: The Pursuit of Objectivity in Science and Public Life. Princeton, NJ: Princeton University Press.

24.

Star

Lampland

(2009) Reckoning with standards. In: Lampland

Star

(eds) Standards and Their Stories. Ithaca, NY: Cornell University Press, pp. 3–24.

25.

Star

Ruhleder

(1996) Steps toward an ecology of infrastructure. Information Systems Research 7(1): 111–134.

26.

Wieczorkowski

(2019) Barriers to using open government data. In: Proceedings of the 3rd international conference on e-commerce, e-business and e-government, Lyon, France, June, pp. 15–20.

27.

Williamson

(2017) Big Data in Education. London: Sage.