This paper offers an empirical analysis of learning management systems (LMS) applying predictive analytics for risk detection in K-12 school education. Conceptually, we explore the promise of predictive analytics to function across various societal domains (e.g. education, predictive policing, digital public service provision) and how LMS configure good educational futures through risk management and mitigation. The starting point for our analysis is the observation that many design features of LMS are risk-related and future-oriented, promising educators and students to achieve success in form of good grades and high graduation rates. Appealing at first sight, this promise of successful educational futures promotes narratives of precision and efficiency inscribed in LMS design (Macgilchrist et al., 2023). Overall, design of educational technologies has been increasingly studied by sociologists of education and critical software and critical data scholars (Decuypere, 2019; Jarke & Macgilchrist, 2021; Selwyn, 2022). So far, this research focuses on the ideologies and imaginaries of technology providers inscribed in the design of educational systems (Macgilchrist, 2019; Manolev et al., 2019; Rahm, 2021; Williamson, 2017). While there are prominent discussions about the role of big tech in shaping the educational domain and the business origin of analytics in education (Davies et al., 2022; Prinsloo, 2019), the centrality of risk and failure in the design of educational technologies has yet to be addressed specifically. Attending empirically to risk prediction in LMS this paper extends on such literature, questioning what and who can be defined as ‘risk’ threatening good educational futures and in which ways.

Narratives about risk and failure, however, are not unique to educational technologies, but are widely discussed in research on predictive policing (Lum & Isaac, 2016; Egbert & Leese, 2021) and digital public service provision (Allhutter et al., 2020; Büchner & Dosdall, 2021). To understand what various ways to define ‘risk’ mean for educational futures, we draw on the concept of “spheres transgression” (Sharon, 2021) to learn about the implications of ‘risk’ inscribed in technologies in other societal domains. Sphere transgression can be understood as an advantageous encroachment of one societal domain into another, making use of distributive capacities of one domain (e.g. big tech) to advance commercially, politically, and socially in the other domains (e.g. education). We argue here that analysing LMS design features we can reconstruct how educational technologies (aim to) reconfigure the organisation of teaching and learning, course design, and interaction between teachers, learners, and administrators by mitigating risks and managing failure.

The (presumable) ability of LMS to produce big quantities of data and to quantify previously unmeasurable societal processes promise educational actors to achieve greater efficiency and more control over the everyday organisation of schooling by managing various educational risks: risk of student drop-out, students failing the course, or graduating from school altogether. To explain how such technological promises are related to actual futures, scholars of technology have connected mundane acts of design, advertising, and negotiation to future-making (Watts, 2015). In these mundane acts, the core characteristics of future-making - anticipation, aspiration, and imagination (Appadurai, 2013) – materialize in form of software design and specific features. To understand and un-make the connection between LMS features of risk prediction and educational futures this paper proposes studying what forms these anticipations, aspirations, and imaginations take in LMS design.

Overall, this paper makes a conceptual and an empirical contribution based on LMS design. Conceptually, following scholars of technology studies concerned with future-making, we shift the analytical focus to the examination of software design from past to the future. Empirically, we analyse risk-related LMS design features. Specifically, we ask how risk is defined in the design of LMS using predictive analytics.

This paper is based on a study of leading international LMS using or providing data for predictive analytics to define ‘risk’ in the context of K-12 education. LMS are designed to generate vast amounts of digital data about their users – both teachers and learners. Some of these data are produced automatically, for example by logging users’ behaviour and interactions within the system such as times spent on certain tasks, number of tasks solved, or courses taken. Other data are the result of both automated and manual labour, for example test scores, teachers’ grades, course attendance data, and uploaded solutions to given tasks. Design features are then understood as relational configurations of use practices, use situations and users that co-construct and co-produce social reality. Using these features and data, LMS configure educational futures through (automated) analysis and prediction. For example, based on current students’ data LMS make predictions about their (likely) success or failure, assigning them higher or lower ‘risk’ scores and providing recommendations to teachers and administrations regarding future pathways of learning. In this paper, we apply the methodology of “feature analysis” (Hasinoff & Bivens, 2021) as a way to reconstruct and analyse how design features frame and configure risk in education. Feature analysis draws on the observation that technologies are designed as solutions to certain problems and aims at identifying how this problem is framed in the design. Feature analysis includes examination of marketing materials (e.g. app descriptions) and graphic user interfaces of the apps. Adopting the feature analysis to the studies of LMS, we examine LMS websites, user handbooks and documentation, available ‘best practice’ cases, and the LMS interface design. We analyse and compare LMS such as Blackboard, Brightspace, Canvas, its Learning, Moodle, Powerschool, and others to identify what these LMS identify as ‘risky’ and which educational actors pose ‘risks’. We qualitatively code the design features these LMS provide to define and predict risks, as well as the kinds of data used to do so (e.g. performance data, interaction log data, sociodemographic data), and actions LMS recommend to educators and students for risk mitigation. By relating these features and data to the three core characteristics of future-making - anticipation, aspiration, and imagination, - we show how LMS design configures educational futures by managing failure and writing out educational indeterminacy and complexity.

This paper aims to show that LMS are designed around risk mitigation and failure management promising various actors to achieve better futures. We propose to analyse risk-related LMS design features considering their future-making capacity and shift analytical attention from studying the ideologies of software providers to the trajectories they draw for further development of education. Using the methodology of ‘feature analysis’, we identify how risk is defined in LMS according to various levels of educational actors posing ‘risks’ – i.e. district, school, student -, and according to what is considered ‘risky’ on some or every of these levels – i.e. failure, inadequacy, inefficiency – which might threaten good school education. We illustrate how the LMS-defined ‘risk’ is bound to in-system interactions (e.g. clicks, uploads, posts) and writes out the contingencies and complexities of teaching and learning processes, foregrounding only certain types of ‘risky’ behaviour over others, taking place outside the LMS. Drawing on the concept of ‘spheres transgression’ we discuss our findings together with insights from research on predictive policing and digital public service provision also concerned with various definitions of risk. So, we show that the LMS definition of ‘risk’ shifts responsibility for ‘risky’ behaviour to individuals, at the same time also foregrounding certain collective actors – schools and districts – particularly prone to include ‘risky’ individuals. Acknowledging similarities in the definitions and implications of technologically-defined ‘risk’ across various societal domains, we discuss what does it mean, when educational technologies become instruments of managing failure to aspire more successful educational futures.

Allhutter, D., Cech, F., Fischer, F., Grill, G., & Mager, A. (2020). Algorithmic Profiling of Job Seekers in Austria: How Austerity Politics Are Made Effective. Frontiers in Big Data, 3. Appadurai, A. (2013). The Future as Cultural Fact: Essays on the Global Condition. Verso. Büchner, S., & Dosdall, H. (2021). Organisation und Algorithmus: Wie algorithmische Kategorien, Vergleiche und Bewertungen durch Organisationen relevant gemacht werden. KZfSS, 73(S1), 333–357. Davies, H., Eynon, R., Komljenovic, J., & Williamson, B. (2022). Investigating the financial power brokers behind EdTech. In S. Livingstone & K. Pothong (Eds.), Education Data Futures: Critical, Regulatory and Practical Reflections (pp. 81–92). Digital Futures Commission, 5Rights Foundation. Decuypere, M. (2019). Researching educational apps: Ecologies, technologies, subjectivities and learning regimes. LMT. Egbert, S., & Leese, M. (2021). Criminal futures: Predictive policing and everyday police work. Hasinoff, A., & Bivens, R. (2021). Feature Analysis: A Method for Analyzing the Role of Ideology in App Design. Journal of Digital Social Research, 3(2). Jarke, J., & Macgilchrist, F. (2021). Dashboard stories: How narratives told by predictive analytics reconfigure roles, risk and sociality in education. BD&S, 8(1). Lum, K., & Isaac, W. (2016). To predict and serve? Significance, 13(5), 14–19. Macgilchrist, F. (2019). Cruel optimism in edtech: When the digital data practices of educational technology providers inadvertently hinder educational equity. LMT, 44(1), 77–86. Macgilchrist, F., Allert, H., Cerratto Pargman, T., & Jarke, J. (2023). Designing Postdigital Futures: Which Designs? Whose Futures? Postdigital Science and Education. Manolev, J., Sullivan, A., & Slee, R. (2019). The datafication of discipline: ClassDojo, surveillance and a performative classroom culture. LMT, 44(1), 36–51. Prinsloo, P. (2019). A social cartography of analytics in education as performative politics. British Journal of Educational Technology, 50(6), 2810–2823. Rahm, L. (2021). Educational imaginaries: Governance at the intersection of technology and education. Journal of Education Policy. Selwyn, N. (2022). Less Work for Teacher? The Ironies of Automated Decision-Making in Schools. In Everyday Automation. Routledge. Sharon, T. (2021). Blind-sided by privacy? Digital contact tracing, the Apple/Google API and big tech’s newfound role as global health policy makers. Ethics and Information Technology, 23(1), 45–57. Watts, L. (2015). Future Archaeology: Re-animating Innovation in the Mobile Telecoms Industry. In A. Herman, Hadlaw, & T. Swiss (Eds.), Theories of the Mobile Internet (pp. 149–169). Routledge. Williamson, B. (2017). Decoding ClassDojo: Psycho-policy, social-emotional learning and persuasive educational technologies. LMT, 42(4), 440–453.