Abstract
Every major technological transformation in education has arrived with the promise of inclusion. Mass schooling, printed textbooks, broadcast media, personal computers, and the internet were each expected to democratize access to learning and opportunity. Yet history offers a sobering lesson: When powerful technologies enter systems already shaped by structural inequality, they often widen divides before they narrow them, if they narrow them at all. Artificial intelligence (AI) now enters education under similarly unequal conditions.
The urgency is difficult to overstate. Globally, more than 70% of children in low- and middle-income countries cannot read a simple text by age ten, a figure that worsened after the COVID-19 pandemic (World Bank 2022). At the same time, nearly one-third of schools worldwide still lack reliable internet connectivity, and in the poorest regions fewer than half of primary schools are connected at all (UNESCO 2023). These disparities shape the baseline conditions into which AI-powered education systems are now being introduced. Against this backdrop, AI is often framed as an automatic equalizer, promising to leapfrog infrastructure constraints, personalizing instruction at scale, and compensating for shortages of trained teachers. This commentary argues that such optimism will be misplaced unless inclusion is deliberately designed into AI-enabled education markets.
AI Adoption and Unequal Value Creation in Education Markets
AI is rapidly reshaping K–12 education systems by automating decision-making at scale. Adaptive tutoring systems personalize content, predictive analytics guide administrative choices, and automated tools are promoted as ways to reduce teacher workload. Once embedded, these systems influence how attention, resources, and expectations are allocated across students, teachers, parents, and administrators in cumulative and path-dependent ways. Evidence from labor markets and organizations shows that AI tends to deliver its greatest productivity gains where complementary investments in skills, infrastructure, and governance are already in place (Acemoglu and Johnson 2023). Consistent with this complementarity logic, randomized evidence from India shows that well-designed, personalized technology-aided instruction can generate substantial test-score gains, with particularly strong relative improvements for initially lower-performing students, but only when the delivery model is operationally robust rather than ad hoc (Muralidharan, Singh, and Ganimian 2019). In weaker institutional environments, the same technologies often substitute for human judgment rather than enhancing it, producing uneven and often unintended consequences. Consider two primary schools adopting the same AI tutoring platform. In a well-resourced district, reliable broadband, trained staff, and ongoing technical support enable the system to personalize instruction effectively, lifting student performance. In a rural school 60 miles away, intermittent connectivity causes the platform to default to generic, grade-level content, while untrained teachers defer to algorithmic recommendations they cannot interpret. Within a single academic year, the same technology widens the gap between these students rather than closing it. Education systems are particularly vulnerable to such divergence because early learning advantages compound across cohorts and over time.
From a marketing perspective, this is not merely a distributional concern; it is a failure of market design. Education is a foundational marketplace in which segmentation, targeting, and value creation determine who develops skills, who comes to see themselves as legitimate participants, and who ultimately enters downstream labor and consumption markets. When AI-enabled education systems systematically serve only institutions and learners best positioned to adopt them, they undermine the logic of effective segmentation by collapsing meaningful heterogeneity into a narrow, advantaged segment. Crucially, these structural failures translate into systematically different experiences: Students in excluded communities encounter AI that erodes trust, undermines belonging, and diminishes self-efficacy, while their advantaged peers experience technology that affirms legitimacy and expands agency (Puntoni et al. 2021).
Such exclusion creates brittle markets that fail to develop broad-based human capital and long-term demand. Reclaiming inclusion is therefore essential to how education markets function, scale, and sustain themselves over time. Classic field experiments in urban India similarly show that technology-supported instruction delivers its largest gains when it is explicitly targeted to students who are lagging behind, reinforcing that inclusion hinges on segmentation and design choices rather than on access alone (Banerjee et al. 2007).
Why AI Can Deepen the Global Education Divide
We identify four predictable points of failure (or theoretical contingencies) that arise when AI adoption proceeds without the conditions needed for inclusive value creation. First, institutional capacity. AI systems depend on stable connectivity, high-quality data, and technical support to generate valid and actionable insights. Where these foundations are weak, algorithmic outputs may be noisy, biased, or misaligned with curricular goals, disadvantaging the very students they are intended to help. UNESCO (2023) documents that deficits in these foundational inputs remain pervasive in low-income contexts, which raises the risk that AI adoption replicates or amplifies existing disparities rather than easing them.
Second, teacher agency. Teachers remain central to learning, particularly for students whose backgrounds and experiences are poorly captured by standardized data. Research on human–AI interaction shows that performance gains are highest when professionals retain discretion to question, override, and adapt algorithmic recommendations (Autor, Mindell, and Reynolds 2022). Yet many AI deployments position teachers as passive intermediaries tasked with executing system directives rather than active partners who shape how technology serves learning. When AI systems constrain judgment rather than support it, they risk deskilling educators and narrowing instructional responsiveness—effects that are more likely to adversely impact marginalized learners. Viewed experientially, teachers in such settings are not merely losing discretion; they are undergoing a shift from professional agency to algorithmic delegation, with consequences for motivation, self-efficacy, and the relational quality of instruction.
Third, cultural and linguistic alignment. Many AI systems are trained on dominant-language data and embed implicit assumptions about learning styles and classroom norms. When AI systems present examples rooted in experiences unfamiliar to students from different cultural or linguistic backgrounds, or when language processing algorithms struggle with nonstandard dialects and multilingual code-switching, the technology communicates symbolic signals about who education is designed for, shaping students’ sense of belonging, recognition, and legitimacy in ways that extend well beyond instructional content. Such misalignment reduces engagement, undermines trust, and can erode the perceived relevance of educational institutions themselves. Over time, these experiences shape aspirations and reinforce perceptions that advanced educational tools are not meant for communities like theirs (Benjamin 2019).
Fourth, governance determines whether exclusionary effects are detected and corrected. AI systems operate through complex, often opaque decision-making processes that can entrench bias without triggering visible failures. Algorithmic harms compound incrementally (misclassifying students, narrowing pathways, reinforcing low expectations) without surfacing until patterns harden across cohorts. Few education systems currently require systematic equity audits of AI tools, despite growing evidence that algorithmic bias can persist unnoticed at scale (Acemoglu et al. 2022). The U.S. Department of Education (2023) reinforces this concern, calling for education-specific guardrails that center educators in decision-making and mandate transparency. Without robust accountability structures, harmful effects diffuse quietly before they are recognized, let alone remedied.
No single actor bears sole responsibility for preventing these outcomes. Platform providers, marketers, school systems, and researchers each shape who benefits from AI at scale. Inclusion fails not because responsibility is absent, but because it is fragmented across actors whose incentives rarely align.
Failure Modes of AI in Education
Table 1 organizes these concerns into a set of failure modes through which AI adoption can unintentionally widen educational inequality. Rather than simply summarizing the argument, the table is intended as a forward-looking research agenda that highlights where empirical inquiry is most urgently needed. It reframes AI adoption as a series of market choices with predictable consequences. Each failure mode generates clear, empirically testable predictions about when AI will exacerbate rather than alleviate inequality. By shifting attention from whether AI “works” in aggregate to how exclusion emerges through specific market mechanisms, the framework invites research that is rigorous, policy-relevant, and central to marketing's role in fostering more inclusive economies. In particular, experiential signals such as disengagement, mistrust, perceived irrelevance, and loss of agency may serve as measurable early warning indicators of exclusion, surfacing harms before they compound across cohorts and harden into structural inequality.
Failure Modes of AI in Education: Research Opportunities for Reclaiming Marketplace Inclusion.
Notes: Illustrative examples are synthesized from recurring patterns documented in recent global reports on AI, education, and digital inequality, including UNESCO's Global Education Monitoring Report (2023), the World Bank's State of Global Learning Poverty (2022), OECD analyses of AI in education systems, and peer-reviewed research on algorithmic bias and human–AI interaction. Examples are representative rather than exhaustive and are used to illustrate mechanisms through which AI adoption may unintentionally reproduce or amplify educational inequality.
Implications for Policy, Pedagogy, and Practice
Preventing AI from deepening the global education divide requires moving beyond celebratory narratives toward more disciplined design. For policymakers, this means aligning AI investments with infrastructure development, teacher training, and governance standards that prioritize inclusion. For educators, it requires preserving professional judgment and embedding AI within pedagogical architectures that recognize and support diverse learners. For marketers and technology providers, it means confronting the tension between scalable growth and inclusive reach, recognizing that markets that exclude at the foundation ultimately undermine their own sustainability. Evidence from experiential learning interventions reinforces this point. Even modest changes in how learning environments are structured can simultaneously improve outcomes and reduce exclusionary beliefs, suggesting that technology must be embedded within inclusive social and pedagogical architectures rather than layered onto unequal systems (Shaik et al. 2026).
In conclusion, AI represents one of the most consequential forces shaping the future of education. The promise is real, as is the peril. Without deliberate attention to institutional capacity, teacher agency, cultural alignment, and accountability, AI risks becoming a powerful engine of stratification, further hardening educational divides that already constrain opportunity for millions of children worldwide. The question is not whether AI will transform education. It already is. The question is whether that transformation will move education markets toward broader participation and long-term vitality or toward deeper inequality and exclusion. Reclaiming marketplace inclusion requires treating AI not as a neutral tool but as a form of market-shaping infrastructure whose design choices will reverberate across generations.
Footnotes
Joint Editors in Chief
Jeremy Kees and Beth Vallen
Special Issue Editors
Samantha N.N. Cross, Rebeca Perren, Eileen Fischer, and Anders Gustafsson
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Data Availability
No data were created or analyzed for this article.