4
Infrastructure
Interoperability, Datafication, and Extrastatecraft
The Gunbarrel Highway is a red dirt desert track snaking through the heart of Australia’s Western Desert. The highway was the first road to cut across the center of Australia from east to west and an important part of Australia’s transport infrastructure. The Gunbarrel Highway has been popularized in Australian culture through the writings of its builder, Len Beadell, and a song by iconic Australian rock band Midnight Oil.1 The highway’s name derives from Beadell’s aim to construct a road that was as straight as a gun barrel, an aim reflecting the desire of the engineer to impose a vision of orderliness on the wild landscape.
The vast and remote Western Desert punctuated by the Gunbarrel Highway may seem a strange starting point for a discussion of data infrastructure in education. The Gunbarrel Highway connects several remote communities from Victory Downs in the Northern Territory to Amata in South Australia and Warburton in Western Australia. Some of these small, remote communities have schools that serve Indigenous people. The network of desert tracks enables families to move between communities at the juncture of three Australian states, and their children may access schools in different jurisdictions at different stages of their schooling. The movement of students across state borders makes it difficult to keep track of students and their attendance and academic records at schools.
There are a lot of moving parts in the production and use of educational data in Australia, which has a federal system of government and an education system governed by the states. In addition, there are also multiple schooling systems within each state, including government and private schools, with the latter being mostly religious and including Catholic, Anglican, “fundamentalist” Christian, and Islamic schools. The need to join up educational data for students, like those along the Gunbarrel Highway, crossing state borders drove an early, and now ongoing, experiment with data sharing and national infrastructure in Australia called the National Schools Interoperability Program (NSIP), the focus of this chapter.
Infrastructures like NSIP are more than the physical hardware—computers and digital networks—required to manage education data. Easterling defines infrastructure as an active form that also includes lesser tangible elements that help to produce new kinds of governing spaces:
The shared standards and ideas that control everything from technical objects to management styles . . . constitute an infrastructure. Far from hidden, infrastructure is now the overt point of contact and access between us all—the rules governing the space of everyday life.2
Infrastructure extends beyond material structures to include software, standards, and ideologies.3 Bowker and coauthors, drawing from Star and Ruhleder, describe infrastructure as “a fundamentally relational concept; it emerges for people in practice, connected to activities and structures. It consists of both static and dynamic elements, each equally important to ensure a functional system.”4 Infrastructure can be understood from this perspective as a dynamic relational form that generates the objects which are the material traces of a functional system. We approach infrastructure as the active form that combines people, networks, algorithms, and computational capacities and that provides the material and nonmaterial support for synthetic governance.
In this chapter, we analyze the development of NSIP and show how building a data infrastructure not only helps to solve logistical problems joining up student data records, but also produces economic and political reconfigurations. We examine how interoperability standards, key to building data infrastructures, create new opportunities for a wider set of commercial and technical actors to play a role in education governance. Based on fieldwork, including interviews, undertaken in 2016–2017 and 2020, we first outline the development of NSIP and the creation of a Learning Services Architecture to connect all Australian schooling jurisdictions. Second, we identify the ways in which infrastructure produces new markets through education governance. We aim to illustrate that developing and implementing data-driven educational infrastructures generates opportunities for commercial providers of data-based products and services to participate in novel ways in education governance. Third, we show how infrastructure, by design, exceeds its original function, using the example of NSIP’s recent role in supporting an online formative assessment initiative, an infrastructure project that involves forms of artificial intelligence (AI).
The Road to a National Data Infrastructure
Interoperability standards specify common data models that enable data to be shared between different applications and systems. Interoperability enables efficient transaction of data and increases the usefulness of data, because it can be exchanged and used across multiple platforms.5 There are considerable benefits in the standardization of data for schools and school systems, including purchasing off-the-shelf applications that can easily be integrated into existing systems.
NSIP was a national project to implement the standards described in the Schools Interoperability Framework (SIF), which specifies common data formats and enables data exchange between school and student information systems and other software. SIF standards are one of a number of standards specifications for educational technologies, with others including IMS Global Learning Tools Interoperability standards and the Shareable Content Object Reference Model.
Microsoft cofounder Bill Gates launched SIF at the U.S. School Administrators’ Annual Conference in February 1999. Gates is no stranger to the world of education policy. The Bill and Melinda Gates Foundation has been an ardent proponent of data-driven education policies and practices through its K-12 Education and Postsecondary Success programs.6 While the Gates Foundation has a philanthropic agenda, this cannot be easily disentangled from its promotion of corporate approaches to philanthropic activity, its network of relationships with corporate actors, and its political lobbying. The standardization and joining up of data systems in education is an agenda that has been pursued by key players in the technology industry since the mid-1990s, and it has continued to gather momentum and political influence. In 2007, the Gates Foundation commissioned a report from the Parthenon Group that reviewed the education data landscape, highlighting the growing market for information management systems and pointing to the commercial opportunities in education.7 Under the auspices of its Institute for Higher Education Policy, Gates aimed to develop a national postsecondary data infrastructure, including linking up data sets across government departments.
In his opening address to the school administrators’ conference, Gates described the need for school districts to develop “digital nervous systems” built on data standards that would provide “a big step forward for both the educational software industry and schools.”8 This initiative was led by Microsoft and supported by eighteen other software companies and the Software and Information Industry Association. The group released the first specification of the SIF standards in November 1999. A more developed specification was released in 2003, and at this time the U.S. Department of Education joined the development efforts. In 2006, a SIF Association was established in the United Kingdom, and in 2009 Australian ministers of education agreed to adopt and develop an Australian specification. In 2015, the Access 4 Learning (A4L) community was established as an overarching body bringing together SIF associations in North America, the United Kingdom, and Australia. The A4L community is a “non-profit collaboration composed of schools, districts, local authorities, states, US and International Ministries of Education, software vendors and consultants who collectively address all aspects of learning information management and access to support learning.”9 The rebadging of the SIF community as the A4L community coincided with an expansion of the focus of these SIF organizations beyond data management to enable “the usage of that data as true learning information for parents, practitioners and learners themselves.”10 A4L argued that its standards provide “the most comprehensive data model and mature infrastructure interoperability framework in use globally in education.”11
SIF and the National Schools Interoperability Program
NSIP was established in 2010 to implement SIF and enable the interoperability of information systems in government and nongovernment schools and school systems across Australia. These school systems previously operated their own information systems with idiosyncratic data formats. NSIP supports a number of projects involving “standards based system integration,” “data sharing between organisations and jurisdictions,” “reuse of infrastructure,” and “data aggregation and synchronisation.”12 The day-to-day technical work of NSIP is undertaken by a relatively small team of approximately ten staff.13
A rationale for ensuring the interoperability of information systems across Australian school systems is the growing federal presence in schooling, which since 2007 has seen the introduction of a national curriculum, a National Assessment Program—Literacy and Numeracy (NAPLAN), and national teacher standards.14 Proponents of the national curriculum argue that students should be able to move between state and territory school systems without missing aspects of their curriculum entitlement. An extension of this argument is that student records should also be transferable between systems to track attendance and achievement, particularly for mobile student groups such as Indigenous students living in remote areas. For example, one of the pilot projects supported by NSIP used SIF standards to develop a system for sharing data between the Western Australia, South Australia, and Northern Territory school systems, three systems that cover Australia’s most remote regions. This project used a unique student identifier and a central application, or agent, to aggregate student data from three jurisdictions and update it on a near real-time basis. Two of the key findings from this project were that to “continue to serve the needs of the Australian education sector, the SIF AU specification requires ongoing development, including regular engagement with local industry and SIF vendors” and that “vendors need access to infrastructure within jurisdictions.”15
Engagement and development across governments, schools, and commercial vendors is a key feature of NSIP. Set up as a joint initiative of federal, state, and territory ministers for education, NSIP operates under the auspices of the Council of Australian Governments Education Council and the Australian Education Senior Officials Committee. In 2017 and 2018, NSIP became subsumed under the umbrella of a ministerial nonprofit company, Education Services Australia (ESA). Each of the state and territory school systems, as well as the Australian government and Catholic and independent school sectors, supports the program. Chief information officers (CIO) from each state and territory education system oversee NSIP’s work as part of a steering group. The technical lead for NSIP described the workings of this hybrid organizational structure:
NSIP is a kind of strange thing. It’s not quite a ministerial company, but . . . we fall under the Australian government. . . . Our steering group is made up of all of the CIOs from the states and territories, and other reps as well, so the Australian government is there. I think as we go forward . . . we’ll see more and more . . . business representation in that group as well.
NSIP is closely aligned with the Australian SIF Association (SIF AU), which has 38 members: 13 governments and government bodies; 9 Catholic and independent school bodies; and 16 commercial vendors. The main product categories offered by these vendors are information management systems, including student information systems (SIS) and timetabling software, but there are a diverse array of companies, including app developers and providers of medical simulation software. Representatives from two of these vendors sit on the Australian SIF Association Management Board, and there are five vendor representatives on the Technical Board. These members play an important role in the development and management of the Australian specification of SIF. An organizational lead for NSIP who acts as a liaison between NSIP, ESA, and the jurisdictions, explains:
We typically do some preliminary modelling, then you put it back to the folks and they see it. It goes through a whole working group process, which is part of the kind of the SIF structure, the standard structure here in Australia. Then if that’s approved from a technical level, it goes to the policy level board, which is [chief information officers] and some selected vendor members, and if that gets the okay it passes on, so it is governed all the way through. There is no opportunity for new pieces to just arise in the standard.
New modalities of digital governance overlap with network governance, which increasingly relies on new linkages between, and movements and analyses of, data. Network governance is also characterized by increasing numbers of “backroom” policy actors such as the technical staff working on developing this digital infrastructure. NSIP provides an example of how technical questions of infrastructure, in this case relating to the SIF standard, allow for a broader range of participants to become involved in education governance, and it illustrates how companies can become new governance players and part of decision-making processes in education, often in quite opaque ways.
The Learning Services Architecture: An Emergent Infrastructure
Data generated across multiple scales and spaces by public and private agencies are becoming increasingly joined up in education systems around the world. The original introduction of the SIF protocol as the basis for a national data-sharing infrastructure in Australia grew into the national Learning Services Architecture (LSA), which was designed “to ensure that education data can be used and re-used consistently in schools, school systems and nationally.”16 The LSA maps out a trajectory toward the establishment of data hubs for all Australian school systems. This model also potentially enfranchises students and schools in relation to data privacy and management through a student data entitlement, and smart contracts that can provide schools and students with control over what data can be used, by whom and for what purposes (Figure 2).17 The LSA was enabled by what the technical lead of NSIP characterized as
a kind of national architecture passion. . . . The nice thing about where we are in Australia is that it actually exists, so this is not something that we’re putting forward as an aspiration to get to in five years’ time; we can talk quite concretely about where everybody is and the state of different pieces, the jurisdictions, the market, the vendors. . . . What the Learning Services Architecture boils down to is that we have an agreed set of patterns across the country, which reflects a change. . . . The whole market forces inside education have shifted. So, there was a time when jurisdictions would see themselves as being the providers of software to schools. You give to schools the systems that they’re going to use, increasingly driven by . . . kind of macroeconomic but also philosophical policy drivers. There’s a move for autonomy in schools, in every sense.
The national LSA conceptualizes NSIP’s various efforts to integrate information systems across schools, school systems, and commercial vendors. Figure 2 represents the role of the SIF standard as information flows (in the middle of the diagram) and the ways that SIF enables interoperability (the common interfaces at the bottom). This figure represents how the establishment of data hubs and information contracts (top left) are creating a new approach to managing data in Australian education.
Similar to those of other countries, Australian public school systems have adopted local school autonomy. The requirements for schools to manage their data have increased and, concomitantly, so has the tendency for systems to manage schools as semiautonomous organizations. The combination of local autonomy and new data accountabilities mean that the use and management of data has become a key part of a hub-and-spoke model through which Australian state and territory education departments govern schools. These networks are similar in the Catholic system and some independent school systems. The LSA creates an architecture that connects these sectors—the hubs and spokes—via a data infrastructure that supports market provision of products and services, with the intention that “schools use the local systems that best meet their local needs” (Figure 2).
Figure 2. Learning Services Architecture. Source: https://www.nsip.edu.au/lsa-2019-roadmap. Figure description.
Figure 3. NSIP information contracts. Source: https://www.nsip.edu.au/lsa-2019-roadmap. Figure description.
One of the key instruments in the LSA, and the underpinning “trusted developer” relationship with hubs, is nationally agreed information contracts that include different aspects of school and system data: enrollment, attendance, assessment, well-being, finance, and learning resources (Figure 3). These contracts cover most of the datafied landscape of education—that is, the areas where significant amounts of administrative and performance data are produced and applied in governance settings.
Importantly, the information contracts address data privacy issues. NSIP does not store any data from the jurisdictions (or hubs), but it does facilitate the movement of data. Along with movement comes concern with privacy issues, a concern necessary as, globally, the development of data infrastructures in education has given rise to resistance based on who has access to data and under what conditions. For example, in the United States the rollout of inBloom, an ambitious centralized platform for data sharing backed by the Bill and Melinda Gates Foundation, collapsed. The failure of this initiative was partly related to privacy concerns and “the combination of the public’s low tolerance for risk and uncertainty and the inBloom initiative’s failure to communicate the benefits of its platform and achieve buy-in from key stakeholders.”18 NSIP is heeding the lessons of inBloom and taking measures to avoid similar challenges to its work.
Making Markets in Australian Schooling
From its early stages, NSIP involved partnerships between government and education technology companies. The NSIP website explains that
in the next 3–5 years [from 2015] the CIOs of all education jurisdictions see a significant shift in their role in the market. This shift will be for education jurisdictions to act as information hubs, exposing student, staff and school data to trusted third party developers, with the expectation that the market will provide products of value to schools that make use of that information.19
NSIP and the implementation of SIF enable the emergence of new network markets. The infrastructure is codeveloped with technology companies (or vendors) that range in size and their level of participation in NSIP, including many smaller companies that were established by ex-education administrators or parents. The implementation of the standard reduces development costs for these companies and provides benefits in relation to the procurement and supply of products and services between schools and vendors.
The NSIP infrastructure produces new relations between jurisdictions, schools, and vendors. Yet if we go back to the launch of SIF by Bill Gates, it seems counterintuitive that the head of a proprietary software company would champion an open standard rather than a proprietary one. Proprietary standards enable companies to “lock in” customers, as occurred with the videotape war between VHS and Betamax formats in the 1980s. Callon argues that lock-in is not simply a deteriorated form of open market in which consumers are captured by companies or products. Instead, by limiting options, lock-in enables calculation and thus helps to constitute and organize markets.
Once the work of standardization (at least partial) of calculating tools is well on its way, each agency is in a position not only to calculate her decision but also, by construction, to include, at least partially, in her calculations the calculations of the other agencies. This integration, which is the material side of what we call anticipation, is far easier when, during the process of market organization, a calculative agency manages to impose directly her instruments and mode of calculation.20
While the imposition of proprietary instruments is a favored lock-in strategy of organizations like Microsoft and Apple, jump-starting the work of standardization is a necessary precursor. From this perspective, we can begin to see the logic of Microsoft championing SIF as an open standard.
In the initial stages of organizing and constituting markets it can be advantageous to join communal efforts to develop open standards. As education technology markets grow, the value of a product depends on how widely it is used. Shapiro and Varian explain that
technologies subject to strong network effects tend to exhibit long lead times followed by explosive growth. The pattern results from positive feedback: as the installed base of users grows, more and more users find adoption worthwhile. Eventually, the product achieves critical mass and takes over the market.21
This positive feedback describes an accelerationist dynamic. Of course, education technology markets do not always grow in this way. For example, the Covid-19 pandemic that closed schools and universities around the world precipitated a rapid move to online platforms, highlighting that the widespread adoption of new technologies can be forced on education institutions, rather than becoming increasingly attractive over time as other market actors adopt them.22
However, setting aside the exceptional circumstances of the pandemic, the market for information management systems in education has developed gradually since SIF was initially specified in 1999. In these circumstances, the development of open standards is a good strategy for growing the total value of the market and potentially sparking explosive growth of positive externalities as the network of users reaches a critical point. This can be a viable strategy even for large players like Microsoft, who can compete in areas such as branding and marketing when the market makes the shift from particular to generic solutions.23 From this perspective, the work of NSIP and the development of data infrastructure in Australian schooling is a strategy for making network markets for data-driven products and services, with benefits for both customers and suppliers.24
Market Integration through Interoperability
Technology companies became involved in the NSIP infrastructure and the LSA due to two potential market benefits. First, Australian state and territory education departments agreed to purchase SIF-compliant student information systems when replacing current systems, and this enabled a change in procurement practices from the central provision of software to more direct relations between schools and vendors, and systems and schools, which increased the market for education technology companies. NSIP oversees the development of a Student Information System Baseline Profile (SBP), which facilitates data exchange between applications and organizations. The SBP uses the SIF data model to specify common data definitions. The Australian SIF Association explains that
the SBP is a breakthrough agreement that defines the relationship between Students, Parents, Teachers, Schools and Classes in a digital machine readable format. It will facilitate the next generation of online services being linked securely into a school’s systems—by reducing the complexity, cost for schools and increasing the ease for vendors.25
The SBP demonstrates the commercial imperative for interoperability standards: reducing the costs for vendors associated with discovering idiosyncratic data formats used by different jurisdictions, which required developing new applications from the ground up for each new customer.
Second, education technology companies can access “dummy” data from schools and jurisdictions through a portal provided by NSIP. This access aids the development of new products. The technical lead explained:
Part of what we do . . . wherever we identify these potential, kind of structural, technical gaps, we provide a whole bunch of tooling that we put out there to market for people to use. . . . And that’s why, if we identify those kind of technical gaps, we try and put tools out there to market; so NSIP doesn’t sell anything, as it were, but we can kind of see those emerging patterns and sort of go, “In the future, you’re going to need to join these things together. Here’s some tools for you to take and use, and do those things.”
As the technical lead indicated, NSIP does not sell a product, but it does join up data sets and make access easier for vendors, creating new conditions for the development of educational technologies.
NSIP provides an “industry forum on technical and interoperability matters for educators and solution providers,” creating an interface between schools and commercial organizations.26 The technical lead explained the attractiveness of this new interface for vendors:
We’ve had a lot of support from the vendor community, because, yes, lock-in is one part of it, but proprietary integration is incredibly painful and expensive. Actually, it means . . . [a vendor’s] product is not economic, so if you get to that level playing field where you can all share, then you have to compete on features, and that’s where they’d rather be, because there’s not much in it in terms of competing over integration. It doesn’t even help you sell it to the school. It’s just a sink cost.
The codevelopment of the SIF standard with commercial actors thus develops potential future benefits for these companies. Prior to the work of NSIP, a school system purchasing a student information system (SIS), for example, would have needed to cover the costs of the supplier discovering its data structures and designing a product that functioned in relation to these structures, within the time frames and budget of the project. If the same school system uses SIF-compliant data structures, then the provider does not need to dedicate the same number of resources to designing bespoke applications. Moreover, the provider can reuse the same applications or software “building blocks” in products for other SIF-compliant school systems, thus widening the market for its products.
The CIO of one of the larger Australian educational technology companies outlined how the SIF standard has created new relationships not only between school systems and companies, but between companies and their competitors:
Anybody who’s ever been involved in software knows that . . . any standardisation process never happens quickly. . . . Having said that, it was a very valuable exercise in taking [SIF from] what was a conceptual data model from something that we could actually all work with and actually use as an agreed standard, . . . That was pretty unusual, to get a bunch of competitors together in the EdTech market to say, “Let’s actually work together for something that doesn’t benefit us commercially directly, but does benefit us in terms of . . . the ability to go out and consume our solution, because the long-term goal is, as a vendor, I could then go to anybody that supports SIF, and theoretically we could just start exchanging data and working together.”
This cooperation between companies may reduce development costs that are passed on to consumers, increasing the competitiveness of all products in the marketplace.
Creating Markets through “Dummy” Data
As mentioned briefly above, the CIOs on the steering group for NSIP designed a model in which vendors can access standardized “dummy” data as a resource for product development. This enables testing to ensure that a particular software package is interoperable before responding to a request for tender. Companies thus stand to gain from a new ability to develop commercial off-the-shelf products or services that can be marketed to, and adapted for, a broader potential customer base. As the technical lead outlined, NSIP specifically aims to perform a governing role while market actors are given new capacities to deliver products to schools:
The role [of NSIP] is one of saying[,] . . . “We can have a much more profound and useful impact as information providers if we focus, actually, on the good governance and management of information, and we leave the actual delivery of the functionality to the market and the things that schools choose,” but it isn’t a free-for-all.
The way in which NSIP enables the integration of different vendors with myriad products into the LSA is thus not only through the standardization of data, but also through providing access to data.
NSIP offered companies development frameworks, training, and a testing service, called the Hub Integration Testing Service (HITS), that enable access to dummy data for product development. HITS is an interface allowing vendors to test whether their software will integrate successfully with existing software used by schools in a particular jurisdiction. HITS fulfills the requirements identified in pilot projects for vendors to have access to jurisdictional infrastructure. The NSIP website outlines that
HITS allows jurisdiction and vendor teams to achieve a level of technical assurance that interoperability will succeed, without having to undertake this discovery process as part of a formal project. HITS itself is a hosted service that allows data interactions and which presents SIF and other suitable endpoints to allow developers to test their system interactions. HITS comes with sample client applications, full developer and administrator documentation and a rich set of credible synthetic school data to make testing and development as meaningful as possible.27
Prior to NSIP, designing for and testing the interoperability of specific products needed to be done on a contract-by-contract basis. The technical lead identified the change HITS enables:
HITS, in effect, is a virtual jurisdiction, so it populates with well-designed dummy data. So as a user we give accounts to any vendor who wants to participate, so again, it’s a free service. Other than asking us for a log-in, [as a vendor] you don’t have to pay anything, and what it does is, when you go in you can say, “Well, I’d like to pretend I’m working with a jurisdiction of 500 schools, I want fifty classes in each, this kind of mixture of teachers.” . . . You can set all these parameters, and in the back we generate all of the data that will then mimic the jurisdiction. Then you’re free to just interact with it.
Put simply, HITS permits vendors to test their software with data that accurately reflects the data structures of a particular school or system. The HITS technology may enable vendors to discover information and develop products before users see a need for them. While tools of this kind create new market opportunities, identifiable personal data is not being exposed to vendors. As noted above, NSIP does not hold any of the data from jurisdictions, and this remains the responsibility of the parties in the information contracts. Rather, NSIP is circumventing the need to discover data systems and structures at the beginning of each contract or project, thus reworking the relationships that constitute policy heterarchies in education. Ball notes that in heterarchies,
different kinds of power relations may exist between the same elements at the same time. Various different kinds of such relationships and asymmetries are currently in play in policy heterarchies—e.g. partnerships, contracts . . . competition, performance management and regulation. . . . Actors and organisations in a heterarchy may play different roles, use different capabilities and exercise different forms of power, at the same time.28
NSIP shows how the development and adoption of interoperability standards can be understood as a market-making process with three important dimensions. First, standardization reduces the time and cost of product development and underpins a shift from particular to generic solutions. Second, the forums required to enact standardization are a state-sponsored space in which suppliers can shape the needs of users, creating demand for products and services that they can profitably provide. Third, the tools for interoperability expose data to vendors and provide them with new resources for product development. In this latter case, we can clearly see how joining up data sets and “liquifying” data creates new sources of value for commercial organizations.
New Commercial Actors and Corporate Social Responsibility
As new commercial actors become involved in building data infrastructures in education, they are subject to political interventions that are sometimes pushing them to emphasize their corporate social responsibility.29 The imperative for companies to narrate their work in moral terms can be demonstrated by revisiting the example of inBloom. As mentioned above, inBloom is another data interoperability initiative that Bill Gates sponsored, through the Bill and Melinda Gates Foundation, and it was launched in the United States in 2013. As Bulger, McCormick, and Pitcan explain, inBloom was “intended to address the challenge of siloed data storage that prevented the interoperability of existing school datasets by introducing shared standards [SIF], an open source platform that would allow local iteration, and district-level user authentication to improve security.”30 However, the initiative generated a powerful backlash of concern about student data privacy and was drawn to a close after one year. InBloom shared similar aspirations to NSIP, and its spectacular failure served as a warning to those involved in similar initiatives elsewhere. There has also been growing public concern and political opposition in response to the level of influence that large and powerful educational publishers, such as Pearson, are gaining in schooling through new data-driven education technologies and associated modes of digital governance.31 These developments are pushing some education technology companies to narrate their business objectives and ethos in ways that differentiate them from the large corporations that have attracted critique and opposition. We present one such narrative, or story of infrastructure, here. This story takes place in 2017.
The Australian education technology market is relatively modest and is populated by many small- to medium-sized companies that have tended to mainly serve one jurisdiction—that is, one state or territory school system—or private schools rather than government schools. As NSIP gains momentum, it may reconfigure this market by opening up growth potential for these companies if they can more easily sell their products into other jurisdictions. However, the value of interoperability standards to commercial vendors is yet to be fully realized in 2017, and those involved in the development of NSIP are not obviously motivated by short-term profits. We spoke to two staff members—a software developer and a CIO—who work at a medium-sized company that has been active in the development of NSIP. Both identify the motivations for this involvement in ways that begin to trouble reductive narratives about commercial interests and technical work.
The software developer works on the technical advisory board for NSIP and explained that, at the time of our interview, there had been “no commercial reason to [participate in NSIP], so why go and invest ten, twenty, thirty thousand dollars developing this stuff if you can’t plug into it?” While the platform may now (as of 2021) provide significant opportunities to “scale up,” this reinforces the point made above that the market was still in the initiation period during which network effects emerge. However, the software developer explained that, despite initial high barriers to entry, their company contributes to pilot projects and participated in technical working groups to help establish the standard. In the process, the company attempts to represent the views of smaller companies that cannot afford to participate in this work, as well as defending the design intent of the standard against attempts to modify it in ways that make it less elegant and functional. The software developer also describes how they help other group members, often from school systems, to better understand how the standard operates and how to configure information systems.
The CIO, another more senior staff member from the same company, explained that the company’s business objectives are different from those of larger multinational companies. Pearson or IBM, for example, actively pursue data-driven technologies that aim to optimize learning and may potentially displace the teacher in the classroom. The CIO explained that the company seeks to stand “out of the way of teachers as much as we can and provide them with the data that they need . . . so that it is complementary to their delivery.” The CIO contrasted this with other prominent corporate visions of technology, including those in which AI is supposed to disrupt traditional teaching and learning modalities.32 The CIO expressed a “hope . . . that doesn’t happen. I think teachers are such a big part of the learning life, I hope that we find a complementary mechanism for teachers . . . to do what they do better.” The CIO continued:
Growing up through an education system you always find . . . people that you look up to not only for the academic thing but for their human qualities, and I think that replacing that teaching by a purely automated thing would mean that people miss out on something that is important. . . . A lot of it is about learning how to be a person.
The CIO presented an unexpected vision of data infrastructure as a complement to liberal humanist education, which is a striking counterpoint to much of the rhetoric about the instrumentality of big data in education and the corporate desire for “disruption.”
From the CIO and the software developer, we heard stories about working closely with a range of partners to undertake careful and collaborative technical work and to develop tools that help teachers do a good job of work. These are stories emerging within a company that was founded by an ex-deputy principal and that employs teachers as part of its workforce. The software developer’s story is framed by a commitment to the profession and its craft, while the CIO’s story is framed by the tradition of valuing teachers as professionals. Neither understands their work in relation to governance, and both invoke desires beyond commercial interests to explain their actions.
Clearly, as we have illustrated above, commercial values and priorities play an important role in the development of a new national data infrastructure for Australian schooling. When we asked whether it is possible that there could be a large player in the Australian student information system, such as North America’s PowerSchool, the CIO responded,
Absolutely. Yeah, I think so. I think the smaller companies were always aspirational to take on other markets, but I think for the reasons I was saying before: it’s not as simple as you might think, and the markets change so much from state to state and country to country, based on how feature-rich you want the product to be, or what value you want the product to deliver to the schools.
The CIO, obviously, would like their company to be that large player, especially because of the CIO’s previous experience with a large higher education technology company. However, the growing education technology market in Australia is being shaped by myriad desires that are part of this infrastructure-in-the-making—desires for profit, for technical quality, and for better education—which do not always align, or at least complicate the other desires and simplistic narratives about commercial actors simply prioritizing profit over other values.
Mission Creep: NSIP, Automation, and Online Assessment
As of 2021, the NSIP data infrastructure provides an updating platform that can be used for a range of purposes that were not necessarily planned for during its initial development. The expanding function of the infrastructure can be characterized as “mission creep.” For example, in March 2018, the Australian government released “Through Growth to Achievement: Report of the Review to Achieve Excellence in Australian Schools.” Known vernacularly as Gonski 2.0, after the head of the review, David Gonski, the report focused on “improving student outcomes,” with recommendation 11 being to “develop a new online and on demand student learning assessment tool based on the Australian Curriculum learning progressions.”33 This recommendation led to establishing the Online Formative Assessment Initiative (OFAI), co-led by ESA, NSIP’s umbrella organization. ESA is a not-for-profit company owned by the education minsters for each state, territory, and the Australian government. It is responsible for providing technology services for education, including the oversight of national infrastructure.
OFAI was established to “provide Australian teachers with innovative assessment solutions that integrate resources, data collection and analytical tools in one ‘ecosystem’ that is easily accessible, interactive and scalable to meet future needs.”34 After a discovery phase with stakeholders, which included NSIP and some of the vendors noted above, OFAI moved into an “alpha” phase in which “tools, resources, protocols and processes will be prototyped and tested, including the preparation of professional development for teachers.”35
Our interest in the relation between NSIP and OFAI is twofold. First, through this relationship the Learning Services Architecture is capable of supporting a broader range of uses. The new Learning Services Architecture 2 (LSA2) is aiming to develop national and state K–12 syllabi as machine-readable assets and to create national data vocabularies. LSA2 also extends the user base of this infrastructure, in which “student, teacher, parent [have] consuming roles.”36 LSA2 involves mission creep due to the intention to manage “the data we need to support education reform.”37 NSIP connects all of the different sections, or “blocks,” of the OFAI’s work, including the “learning progress tracker, online test creation and delivery, and suggestion engine” (Figure 4). As the NSIP technical lead explained:
We [NSIP] are the OTF, as it were, the open technology framework, which in effect is the linking piece between all those blocks . . . between all the parties. We kind of touch everyone. We don’t get too far into the detail of what each one of those blocks does, I guess it’s fair to say.
In practice, NSIP’s data infrastructure extends beyond the implementation of SIF standards. A discovery stage report notes that the OFAI will be underpinned by an open technology framework (OTF), including student single sign-on capability. While it may appear to be a minor detail, the report on the OTF recommends that this sign-on “use SIF-AU where available” and “use other widely used APIs and standards where available, including Google Classroom.”38 Thus, while SIF is still preferred, the infrastructure will enable the participation of other vendors, such as Google, to support OFAI.
Figure 4. Organizing “blocks” of OFAI. Source: https://www.lpofai.edu.au/. Figure description.
Second, establishing OFAI shows how data infrastructures in education are the necessary precondition, and enabler, of new forms of automated decision-making in education. OFAI explicitly focuses on AI in both online test creation and delivery and the provision of a “suggestion engine” that will be used to rank suggestions for learning needs. As the technical lead stated,
We use AI in these layers particularly to assist with alignment of inputs to the learning progressions. And I won’t make any sort of outrageous claims about the way we do it. It’s simple Bayesian classification stuff, as it so often is.
AI, in early and simple forms, is becoming part of the networked governance enabled by NSIP’s data infrastructure. The new online assessment initiative does not radically change what NSIP does but demonstrates that the infrastructure expands beyond being a method for data standardization and interoperability across school systems and vendors. The work being undertaken as part of the OFAI illustrates that the infrastructure is infinitely expandable. capable of embracing all areas of education, from school-based administration, to national assessment, to the introduction of AI into schooling systems.
Infrastructure, Governance, and Enabling Automation
In this chapter, we have illustrated how building data infrastructures creates new opportunities for a wider set of commercial and technical actors to play a role in education governance through mundane technical work such as developing and implementing standards. Data infrastructure, we posit, is a key ingredient in the emergence of synthetic governance as it generates technical, economic, and political reconfigurations. To conclude this chapter, we will discuss three key points of this reconfiguration regarding infrastructure and markets, governance, and the enabling conditions for automation.
Infrastructure and Market-Making
NSIP is a key, yet hidden, actor in the making of new markets for education technology in Australian schooling. The development of data infrastructure involves the specification of standards that become part of its “active form” or disposition. Standardization helps to frame and stabilize the various objects or goods that are exchanged in markets. Following Callon, markets can be understood as devices that coordinate the activities of calculating agencies. These agencies are more than an individual person and include “the material and devices . . . that give his or her actions a shape.”39 Neither the market nor homo economicus preexists the tools and social relations that bring these entities into being as a mutually constitutive process and organization of relations between agencies. These tools include accountancy practices, economic rationalities, and various measurement tools, as well as standards that help to disentangle and frame objects of market exchange.
The emergence of calculative agencies requires framing. Callon argues that framing “allows for the definition of objects, goods and merchandise which are perfectly identifiable. . . . It is owing to this framing that the market can exist and that distinct agents and distinct goods can be brought into play.”40 Çalışkan and Callon argue that such “disentanglement is more stable . . . when a commodity has undergone specific processes of standardization that transform it into an entity described in both abstract and precise terms, certified and guaranteed by a series of textual and material devices.”41 From this perspective, markets are continually performed through processes of disentanglement and framing. The state plays an essential role in the constitution of markets through these processes by funding infrastructure, establishing regulatory frameworks, and sponsoring standard-setting forums, as illustrated above. Standards play an important role in framing goods, and the standardization of quantities, price, and currency creates the very conditions that “make modern markets possible.”42
Governance and Infrastructure
Infrastructure not only creates markets, but it also supports and creates new modes of governance. Easterling argues,
Infrastructure space has become a medium of information. The information resides in invisible, powerful activities that determine how objects and content are organized and circulated. Infrastructure space, with the power and currency of software, is an operating system.43
The analogy with an operating system underlines the active nature of infrastructure. For example, computers and cables are object forms, whereas standards and codes are active forms. Easterling emphasizes active forms, arguing that “infrastructure space is a form, but not like a building is a form; it is an updating platform unfolding in time to handle new circumstances.”44 Data infrastructures can thus be understood as active and changing platforms for storing, sharing, and consuming data across networked technologies. This definition emphasizes the issue of interoperability—that is, the capacity for the platform to operate across different material supports (local or virtual machines, for example) and to share data between different systems. In other words, “the shared standards and ideas that control everything from technical objects to management styles . . . constitute an infrastructure.”45
NSIP created new relationships between private and public actors as the state built public infrastructure that acts as a quasi-incubator of the education technology market. NSIP is an example of “extrastatecraft,” where the role of education technology companies in data infrastructures is “a largely hidden dimension of the privatization and commercialization of government schooling systems.”46 The infrastructure is a response to, but also further facilitates and intensifies, longer-standing trends toward devolution and local school management. Additionally, the governance relationships of NSIP are formed around data hubs that replicate the existing territories of educational governance (state authority over education) and bring multiple new actors into a new infrastructure space (cross-sector, national education technology markets), thereby reforming jurisdictional (geographical) spaces and creating new connections between different educational sectors and actors. The enabling of new forms of data management by NSIP is thus helping to reconfigure education governance in Australia.
Infrastructure and the Enabling Conditions for Automation
Finally, we can see that the example of NSIP’s role in OFAI and the emergent use of AI in education governance creates conditions for new problematizations. On the one hand, OFAI is the outcome of a policy decision—Gonski 2.0—and NSIP enables the implementation of this policy. On the other hand, and unsettling common perspectives on network governance, infrastructure becomes the necessary precondition for OFAI because it requires a national infrastructure that is standardized and interoperable. As a result, the NSIP infrastructure exceeds its function as a primarily administrative infrastructure for linking student information systems and data. The updating functionality of this infrastructure is the outcome of a policy decision and the capacities provided by the active form. NSIP, thus, contributes to creating the conditions of possibility for new types of algorithmic governance in education, and highlights the relatively mundane ways in which AI is being introduced. These developments demand a new problematization of technology, specifically AI, and its relationship to education. As Mackenzie suggests,
Everybody has heard that technology will make the future better. Everyone had heard that it could make it a lot worse. The problem is to figure out what scope or capacity there is to collectively manoeuvre between these two somewhat bloodless abstract possibilities.47
The NSIP infrastructure emerged in part as a response to a problem created by the establishment of an earlier infrastructure—desert tracks that joined up communities in Australia’s remote Western Desert. The work of NSIP now generates conditions in which new developments can occur, including the automation of aspects of teaching, learning, and assessment. In the next chapter, we explore examples of these new technologies that are already being implemented in schools, with a specific focus on facial recognition. The challenge that must be confronted is not whether these developments are “good” or “bad” for schools, and for the teachers and students who are increasingly positioned as users, but what scope for maneuver there is within these unfolding platforms.