Across 2015-2025, educational technology has shifted from peripheral classroom “tools” to infrastructures that classify, predict, monitor, and intervene in learning at scale (Williamson, Komljenovic, & Gulson, 2024). This includes learning analytics and datafication; generative AI, tutoring and recommender systems; platformization and cloud “stacks”; brokered procurement and “evidence” infrastructures; and, increasingly, technologies oriented to minds and bodies, emotion/attention sensing, neurotechnologies, and educational genomics. These developments are routinely framed as inclusion-enhancing, promising personalization, early identification, efficient support, and “evidence-based” decision-making. However, they may also reproduce exclusion through automated categorization, surveillance and psychodatafication, market dependency, and bio-determinist imaginaries of ability (Perrotta & Selwyn, 2020).

This proposal investigates how exclusion is produced by new trends in educational technology even when framed as inclusion, through a systematic review of empirical research published between 2015 and 2025. The review is theoretically anchored in the conceptual framework of “zombie education” (Dovigo, 2024), which approaches inclusive education as a contested “boundary object” stabilized by institutional routines and classificatory infrastructures that can become normalized, black-boxed, and politically resilient. Within this framework, exclusion is not merely a failure of inclusion; it is often generated through inclusion itself via two intertwined processes:

• Immunization (managed inclusion): difference is “welcomed” by being broken into administratively manageable “dividuals” through labels, diagnostics, screening, and risk apparatuses, protecting institutional norms while performing inclusion (Esposito, 2011).
• Burnout (over-inclusion): inclusion is amplified through incessant measurement, optimization, ranking, and repetition until difference becomes devalued and the system exhausts learners and educators (Han, 2015).

We extend these concepts from SEN policy analysis to contemporary edtech by treating technologies as apparatuses/assemblages creating configurations of platforms, data practices, policies, procurement routines, markets, and discourses that actively shape educational subjectivities and possibilities. This includes “nudge” logics and micro-interventions that subtly steer conduct through data-driven personalization (Braidotti, 2019), as well as algorithmic governance aligned with broader “societies of control” (Deleuze, 1992).

Research questions:
- RQ1: Which edtech trends (2015-2025) are most frequently linked to inclusion/equity claims, and how is “inclusion” defined and operationalized?
- RQ2: Through what sociotechnical mechanisms do these trends produce exclusionary pressures (immunization and/or burnout), and for whom (e.g., SEN/disability, minoritized learners, low-income communities)?
- RQ3: What inclusion-enhancing opportunities are evidenced, and what enabling conditions (design, pedagogy, governance, resourcing) recur?

The objective is a critical synthesis showing how “inclusive” edtech can become zombie inclusion, whit inclusion animated in language and policy, while exclusion is reproduced through infrastructures of classification, evidence-making, and political economy.

We conducted a systematic review of literature published 2015–2025, reported using PRISMA 2020 (Page et al., 2021). Searches were conducted in ERIC, Scopus, Web of Science, PsycINFO, and PubMed (to capture neurotechnology/genomics-related educational research). We included peer-reviewed empirical studies (qualitative, quantitative, mixed-methods) and high-relevance review studies. In addition, we included selected grey literature and policy-facing reports where integral to understanding edtech trends (e.g., platform governance, brokerage/evidence infrastructures, neuro/genomics governance). Inclusion criteria: (a) educational setting (K-12, higher education, adult/vocational); (b) a defined edtech trend (AI/generative AI; learning analytics; platform/cloud infrastructures; brokered procurement/evidence systems; surveillance/SEL/psychodata; neurotechnology; educational genomics); and (c) explicit engagement with inclusion/exclusion, equity, SEN/disability, marginalization, or distributive effects, informed by intersectional framings of inclusion (Thomas & Macnab, 2022).
Exclusion criteria: purely technical papers without educational implications; studies outside 2015–2025; items lacking sufficient methodological detail. Screening and reliability: two-stage screening (title/abstract; full text). A subset was dual-screened; disagreements were resolved through documented decision rules; a PRISMA flow diagram was produced. Quality appraisal: MMAT (2018) for heterogeneous evidence, supplemented by CASP for qualitative studies when appropriate (Hong et al., 2018; CASP, 2018). Data extraction: technology type; educational sector; governance/business model (platform/cloud; vendor; brokerage); populations and learner groups; operational definition of inclusion; outcomes/claims; risks/harms; and forms of classification/measurement (screening, prediction, profiling, ranking), including “nudge” mechanisms where present (Decuypere & Harttong, 2023). Synthesis: thematic synthesis structured by immunization and burnout mechanisms, with an additional mapping of assemblage features (discourses, institutions, laws, platforms, data flows, markets) to show how exclusion is co-produced across sociotechnical configurations.

The review produced an evidence-informed risk-opportunity map indicating when edtech supports inclusion and when it generates zombie inclusion, with inclusive intent and discourse sustained while exclusion is operationally reproduced. Inclusion-enhancing opportunities (conditional rather than automatic) cluster around accessibility and assistive supports; translation and multimodal scaffolds; targeted supports embedded in inclusive pedagogy and relational classroom practices; and participatory governance/design practices that de-black-box systems and redistribute decision-making power. Exclusion-producing risks cluster around two main dynamics: Immunization via datafied classification: AI and analytics, SEL/psychodata, attention/emotion sensing, neuro-measures, and educational genomics may intensify labeling, prediction, and sorting, reframing inequality as individual (or biological) deficit and legitimizing differential expectations and allocations (Morris et al., 2024). Burnout via infrastructural optimization: platform/cloud stacks, brokered evidence regimes, and investor-driven scaling may amplify metric saturation and “evidence” performativity, increase compliance/data labor, narrow educational aims to measurable proxies, and divert resources from relational and structural inclusion work, especially under austerity conditions (Mau, 2019). The “hype–uncertainty” dynamics of algorithmic surveillance can further accelerate adoption while muting contestability (Mackenzie, 2024), deepening the conditions for burnout and democratic deficit (Foucault, 2008). The principal contribution is a conceptual model of smart exclusion, showing how emerging edtech extends immunization and burnout through sociotechnical assemblages, aligned with critical perspectives on what counts as “critical” in edtech research (Selwyn et al., 2021).

Braidotti, R. (2019). Posthuman knowledge. Polity Press. Decuypere, M., Harttong, S. (2023) Edunudge. Learning, Media and Technology 48(1): 138–152. Deleuze, G. (1992). Postscript on the societies of control. Pourparlers. Editions Minuit. Dovigo, F. (2024). Towards a Zombie Theory of Inclusive Education: A Discourse Analysis of Special Educational Needs Policies in Five European Countries. European Journal of Inclusive Education, 3(2). Esposito, R. (2011). Immunitas: the protection and negation of life. Polity Press. Foucault, M. (2008). The courage of truth: The government of self and others II. Lectures at the College de France, 1983–1984. Palgrave MacMillan. Han, B. C. (2015). The burnout society. Stanford University Press. Hong, Q. N., et al. (2018). Mixed Methods Appraisal Tool (MMAT), Version 2018: User guide.
Critical Appraisal Skills Programme (CASP). (2018). CASP qualitative checklist.
Mackenzie, A. (2024). The secret life of data: Navigating Hype and Uncertainty in the Age of Algorithmic Surveillance. The MIT Press. Mau, S. (2019). The metric society: On the quantification of the social. Polity Press.
Morris, T. T., von Hinke, S., Pike, L., Ingram, N. R., Davey Smith, G., Munafò, M. R., & Davies, N. M. (2024). Implications of the genomic revolution for education research and policy. British Educational Research Journal, 50(3), 923–943. Page, M. J., et al. (2021). The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ, 372, n71. Perrotta, C., & Selwyn, N. (2020). Deep learning goes to school: Toward a relational understanding of AI in education. Learning, Media and Technology, 45(3), 251–269. Selwyn, N., Hillman, T., Bergviken Rensfeldt, A., & Perrotta, C. (2021). What is ‘critical’ in critical studies of edtech? Three responses. Learning, Media and Technology, 46(3), 243–249. Thomas, G., & Macnab, N. (2022). Intersectionality, diversity, community and inclusion: untangling the knots. International Journal of Inclusive Education, 26(3), 227-244. Williamson, B., Komljenovic, J., & Gulson, K. N. (Eds.). (2024). World yearbook of education 2024: Digitalisation of education in the era of algorithms, automation and artificial intelligence. Routledge.