In April 2016 the Education Endowment Foundation launched the Families of Schools Database, a searchable database that allows any school in England to be compared with statistically-similar institutions. At about the same time, the Learning Analytics and Knowledge 2016 conference was taking place in Edinburgh, focused on the latest technical developments and philosophical and ethical implications of data mining learners and ‘algorithmic accountability.’ The current development of school comparison websites like the Families of Schools Database and the rapid growth of the learning analytics field point to the increasingly fine-grained, detailed and close-up nature of educational data collection, calculation and circulation–that is, they offer a kind of numerical and ‘intimate’ analytics of education.
One way of approaching these school comparison databases and learning analytics platforms is through Kristin Asdal’s notion of ‘accounting intimacy.’ According to Asdal, practices of calculation are increasingly moving away from bureaucratic practices enacted in distant ‘centres of calculation’ to much more ‘intimate’ calculative practices that are enacted in situ, close to the action they measure. Intimacy also implies close relationship, and practices of calculative or accounting intimacy can also be understood in terms of how numbers and numerical presentations of data can be used to build intimate relationships between different actors.
Radhika Gorur has adapted Asdal’s ideas to suggest that more ‘intimate accounting’ is increasingly occurring in education. Drawing on the example of the school comparison site MySchool in Australia, she argues that:
The public, especially parents, was exhorted to make itself familiar–intimate–with the school by studying the wealth of detail about each school that was on My School. The idea was that, armed with intimate knowledge of their child’s school, parents could exert pressure on schools to perform well and get the best outcomes for their children. Not only did My School become a technology through which the government entered intimate spaces of schools, schools themselves entered intimate spaces of living rooms and kitchens through discussions between parents.
Through these techniques, schools could become available for the intimate scrutiny of the government as well as by parents.
The Families of Schools Database, like Australia’s MySchool, involves schools in providing highly intimate details—in the form of numbers—that can then be presented to the general public. These public databases allow the school to be known and discussed, as Gorur argues, in the intimate spaces of the home—as well as involving school leaders in the intimate accounting and disclosure of their institution’s performance according to various criteria. One aim of the Families of Schools Database is to enable statistically-similar schools to identify each other and then collaborate to overcome shared problems. An intimate knowledge of other institutions is required to facilitate such collaboration (thought it might also motivate competition). While school data certainly are collected together and transported to distant centres of calculation to allow the compilation of such databases, a certain demand is placed on institutions to present themselves in terms of an intimate account, and ultimately to share that account as a means towards possible collaboration with their numerical neighbours.
Following Ingmar Lippert we might say that such practices of intimate accounting configure the school environment as a ‘datascape,’ one whose existence in organizational reality is achieved through the calculative practices that make it ‘accountable.’ By configuring the school environment as a ‘dataspace,’ as Lippert argues, ‘reality is enacted’ as its intimate details are projected as a stabilized numerical account. Databases such as the Families of Schools Database might therefore be understood as intimate datascapes, where schools’ data are disclosed with the aim of building close relationships with parents and other institutions, whilst also becoming more visible to government.
When it comes to learning analytics, the level of intimate accounting is increased even further. With such systems comes the technological ambition to know the microscopically intimate details of the individual learner. Major learning analytics platform providers such as Knewton claim to collect literally millions of data points about millions of users to amass huge big datasets that can be used for the automatic analysis of learning progress and performance.
For Knewton, the value of big data in education specifically is that it consists of ‘data that reflects cognition’—that is, vast quantities of ‘meaningful data’ recorded during student activity ‘that can be harnessed continuously to power personalized learning for each individual.’ The collection and analysis of this ‘data that reflects cognition’ is a sophisticated technical and methodological accomplishment. As stated in documentation on the Knewton website:
The Knewton platform consolidates data science, statistics, psychometrics, content graphing, machine learning, tagging, and infrastructure in one place in order to enable personalization at massive scale. … Using advanced data science and machine learning, Knewton’s sophisticated technology identifies, on a real-time basis, each student’s strengths, weaknesses, and learning style. In this way, the Knewton platform is able to take the combined data of millions of other students to help each student learn every single concept he or she ever encounters.
The analytics methods behind Knewton include Item-Response Theory, Probabilistic Graphic Models, and Hierarchical Agglomerative Clustering, as well as ‘sophisticated algorithms to recommend the perfect activity for each student, constantly.’
What a learning analytics platform like Knewton appears to promise is a highly intimate and real-time analytics of the very cognition of the individual, mediated through particular technical methods for making the individual known and measurable. Again, as with the Families of Schools Database, it is clear that the data are being collected and transported to distant centres of calculation—namely Knewton’s vast servers—but the speed of this transportation has been accelerated massively as well as being automated. A vast new datascape of cognition–amassed methodologically according to the psychometric assumptions underlying Item-Response Theory et al–is emerging from such calculative practices.
Moreover, because Knewton’s platform is adaptive, it not only collects and analyses student data, but actively adapts to their performance so that each individual experiences a different ‘personalized’ pathway through learning content, as determined by machine learning algorithms. Such algorithms have the capacity to predict students’ probable future progress through predictive analytics processes, and then, in the form of prescriptive analytics, to personalize their access to knowledge through modularized connections that has been deemed appropriate by the algorithm. To give a sense of this, in Knewton’s documentation, it is stated that all content in the platform is:
linked by the Knewton knowledge graph, a cross-disciplinary graph of academic concepts. The knowledge graph takes into account these concepts, defined by sets of content and the relationships between those concepts. Knewton recommendations steer students on personalized and even cross-disciplinary paths on the knowledge graph towards ultimate learning objectives based on both what they know and how they learn.
The Knewton platform’s ‘knowledge graph’ treats knowledge in terms of discrete modules of content that can be linked together to produce differently connected personalized pathways.
In this sense, knowledge is treated in terms of a network of individual nodes with myriad possible lines of connection, and the Knewton platform ‘refines recommendations through network effects that harness the power of all the data collected for all students to optimize learning for each individual student.’ For Knewton, knowledge is nodal like a complex digital network, and constantly being refined as machine learning algorithms learn from observing large numbers of students engaging with it: ‘The more students who use the Knewton platform, the more refined the relationships between content and concepts and the more precise the recommendations delivered through the knowledge graph.’ In other words Knewton is developing new kinds of intimacies between units of content and concepts, as well as identifying recommendations for students that are based on an assessment of the optimal relationship between the individual learner and the individual content item. The Knewton knowledge graph ultimately consists of networked data that reflects content, and data that reflects cognition, and it is constantly analyzing these data to find best fits, clusters, connections and relationships–or numerical intimacies in the datascape of content and cognition.
Real-time intimate action
What I am briefly trying to suggest here is that a kind of automated real-time intimate accounting at the level of the individual is occurring with these learning analytics platforms. Such platforms both govern learners at a distance—through transporting their data for collection and processing via data servers and storage facilities—but also up-close, intimately and immediately, through real-time adaptivity and personalized prescriptive analytics.
Whereas the Families of Schools Database and MySchool involve more intimate accounting among human actors mediated through public databases, however, the intimate action of learning analytics is algorithmic and subject to machine learning processes. The ambition of Knewton, and other learning analytics platform providers, is nothing less than an intimate account of the individual, which can then be analyzed as points in a vast networked datascape of content and cognition of others to ‘optimize learning’–and in this sense it instantiates a distinctive form of real-time intimate action that is targeted at individual improvement at the level of cognition itself.
Kristin Asdal suggests that intimate accounting involves the ways that calculative practices associated with ‘the office’ become implanted in ‘the factory’–that is, bureaucratic practices of distant data collection and calculation are displaced by practices of enumeration that are enacted much more closely to the measurable action. Schools are now increasingly involved in their own practices of institutional intimate accounting and the production of the school environment as a datascape. The proliferation of learning analytics platforms brings intimate accounting into the everyday life and learning of the individual, with algorithms (and the methodologies that underpin them) designed to provide both an intimate account of the individual–as data that reflects cognition–and to undertake intimate action in the shape of prescriptive analytics and automatically personalized learning pathways that might shape the individual as a cognizing subject.