A study in Frontiers in Education argues that universities have narrowed their response to artificial intelligence in the classroom to a single, insufficient concern: catching students who cheat.
The paper, by Dr. Kelechi Ekuma of the University of Manchester's Global Development Institute, targets development studies programmes, those that train graduates for careers in international development policy and humanitarian work. But its central argument applies more broadly. Ekuma contends that universities have focused on AI as an academic integrity problem while largely ignoring what the technology means for the kinds of graduates they produce and the jobs those graduates will enter.
"AI and automation should be conceptualized not merely as new technologies entering higher education, but as structuring conditions that are reshaping the epistemic, pedagogic, and professional environment within which development studies operate," he wrote.
Published July 2, the paper arrives as the gap between AI adoption and policy at universities has become increasingly visible. A 2026 global report on AI in higher education found that the gap between AI use and proper governance "is still a major challenge." Surveys from Middlebury College and Yale University found that more than 80 percent of students had used generative AI for academic work within two years of ChatGPT's launch in 2022. A HEPI-Kortext survey separately found that 58 percent of students use AI to explain concepts. Those numbers reflect student behavior that is already far ahead of institutional guidance.
Plagiarism detection is not an education strategy
The early focus on plagiarism, Ekuma writes, "risks reducing a profound educational transformation to a problem of surveillance and compliance." Researchers Kofinas et al. (2025), cited in the paper, show that authentic assessment faces real vulnerabilities when generative AI can produce plausible text rapidly and at scale. The paper argues that returning to invigilated exams or relying on AI detection tools is not an adequate response. Literature Ekuma cites treats detection tools as unreliable as a foundation for future assessment systems.
What the paper proposes instead is teaching students what Ekuma describes as "critical AI literacy." This is not a skill in using AI tools. It is the capacity to question what AI systems cannot do, and to situate those limitations within the political and institutional context that shapes whose knowledge the systems actually represent.
The urgency behind this argument is grounded partly in what AI is already doing in development contexts outside the classroom. AI classification systems determine welfare eligibility. Predictive systems manage agricultural advice and credit scoring in settings where institutional oversight is already limited. These are not futuristic scenarios. They are conditions that development graduates will encounter in their first jobs.
"This challenge is especially urgent because AI and automation now cut across domains that have long been central to development scholarship," Ekuma wrote. "They are being embedded into public administration, welfare targeting, agriculture, finance, health, education, identity systems, humanitarian response, and labour management."
AI is already inside university operations
The classroom debate is unfolding inside universities that have already integrated AI across multiple administrative functions. A 2025 Brandon Hall Group study, cited in U.S. enrollment automation research, found that automated systems reduced application processing times by 40 percent and recovered up to 60 percent of applications that might have been abandoned due to administrative delays. Schools now use predictive systems to flag students who may be at risk of dropping out, based on academic performance and advisor meeting records.
In India, an EY-Parthenon and FICCI report found that 53 percent of universities use generative AI to produce learning materials, with 38 percent having introduced AI into grading processes. The pattern points to AI adoption that has moved past the experimental phase inside many institutions, well before those institutions established governance frameworks to guide it.
Ekuma's paper frames the mismatch between operational adoption and pedagogic response as a disciplinary failure. Technology has long appeared in development studies curricula as a sectoral theme, digital inclusion, e-governance, ICT for development. That framing is no longer adequate, he argues, because AI is no longer a topic within development. It is increasingly part of the infrastructure through which development decisions get made.
Government action has accelerated but curriculum has not kept pace
Governments have moved on AI education faster than universities have updated their programmes. In April, U.S. President Donald Trump signed an executive order establishing a White House Task Force on AI Education and directing federal agencies to expand AI programs for students and teachers. The U.S. Department of Labor also launched an AI apprenticeship portal, with training targeted at sectors including finance and healthcare. Google's philanthropic arm announced a $2 million initiative with the Sundance Institute to train more than 100,000 artists on AI tools.
Mississippi College School of Law went further by requiring first-year students to complete AI coursework focused on understanding the technology and verifying its outputs. Ekuma's paper calls for something similar, but embedded across full degree programmes rather than confined to a single required course.
Assessment design sits at the center of that argument. The paper proposes that written work be complemented by oral defence of written analysis and structured critique of AI-generated outputs, forms of evidence that require students to demonstrate reasoning rather than just product. Ekuma is direct about what drives that proposal.
"This does not mean every module must become a module on AI," he wrote. "It means that existing modules should reconsider how AI reconfigures the issues they already teach. In this sense, curriculum integration should be additive in scope but transformative in implication."
What the gap between adoption and critical training costs graduates
The employability dimension of Ekuma's argument draws on research showing that graduates are evaluated by employers not only for what they know, but for how they interpret and make decisions under uncertainty. Anderson and Tomlinson (2021) show that employer assessment of graduates is relational and goes well beyond formal qualifications.
In AI-mediated workplaces, that relational dimension does not diminish. Tasks that are codifiable and heavily text-based may be automated or augmented. The value of a human graduate shifts toward contextual interpretation and ethical reasoning, qualities Ekuma argues development-oriented programmes are well placed to develop, but only if they assess for them deliberately.
For universities in the Global South, the challenge carries additional weight. Where bandwidth costs and local-language AI coverage are uneven, an AI-responsive curriculum cannot replicate models designed for well-resourced institutions. Ekuma argues that in those settings, questions about data ownership and whether institutions have the capacity to exercise meaningful oversight of AI-mediated decisions should come before any discussion of advanced tool use.
Ekuma's paper was published as a conceptual and integrative review in the journal's Higher Education section on July 2, 2026. The author acknowledged in the paper that AI-supported tools were used during drafting and language refinement, a disclosure that reflects the transparent AI-use standards the paper argues universities should require of students.
Tanya Petrusenko
Disclaimer: All materials on this site are for informational purposes only. None of the material should be interpreted as investment advice. Please note that, despite the nature of much of the material created and hosted on this website, HODL FM operates as a media and informational platform, not a provider of financial advisory services. The opinions of authors and other contributors are their own and should not be taken as financial advice. If you require advice, HODL FM strongly recommends contacting a qualified industry professional.
