How AI Can Strengthen Inclusive Education Across the Global South

Rural universities are being asked to prepare teachers for increasingly diverse classrooms without the infrastructure, specialist expertise or digital support available to better-resourced institutions. Artificial intelligence may help close some of those gaps, but only when universities invest as seriously in people, policy and access as they do in technology.

A new study "AI-Enabled Capacity Building for Special Education Pre-Service Teachers: A Systematic Scoping Review of the Global South with Implications for South African Rural Universities," published in the journal Trends in Higher Education, examines this challenge. Authored by Mamochana Anacletta Ramatea and Sumeshni Govender of the University of Zululand, the study maps how AI is being used to strengthen the skills of future special-education teachers, and why its benefits remain difficult to realize in under-resourced institutions. The authors reviewed 50 peer-reviewed studies selected from 2,040 records published between 2010 and 2025.

The study's message is both optimistic and cautionary. AI can improve lesson planning, assessment, reflective practice, simulation and data-informed teaching. But where internet access is unreliable, lecturers lack training and institutional policy is weak, AI may deepen rather than reduce educational inequality.

AI Can Extend Teacher Capability, Not Replace It

The key finding is that AI's value in special education lies in helping teachers respond to learner diversity. Adaptive systems, intelligent tutoring tools, speech-recognition applications and AI-assisted analytics can support differentiated instruction, personalized learning and more responsive assessment.

For pre-service teachers, these tools can make training more practical. AI-enabled simulations and case-based exercises can expose trainees to complex classroom situations before they enter schools. Digital dashboards can help them analyse learner responses, identify patterns and reflect on whether their teaching strategies are working.

In special education, where teachers must often adapt materials and methods to individual cognitive, behavioural or communication needs, this kind of guided practice can be particularly valuable.

However, the review rejects the idea that technology can substitute for professional judgment. Special education depends heavily on empathy, context, cultural understanding and the ability to interpret behaviour that may not fit standardized categories. AI can generate options, flag patterns or suggest resources, but teachers must decide whether those outputs are appropriate, accurate and ethical.

Poorly designed training can create superficial competence. A trainee may learn how to prompt a generative AI system without learning how to detect bias, protect sensitive data or adapt material for local languages and cultural settings.

The review treats AI literacy as a combination of technical, pedagogical and ethical capability, not simply the ability to operate software. Its value depends less on the sophistication of a tool than on the teacher's ability to question, contextualize and apply its output responsibly.

Rural Universities Face a Capacity Trap, Not Just a Connectivity Gap

The largest obstacle is often described as poor internet access, but the study shows that the problem is broader. Rural universities may face unreliable connectivity, limited devices, insufficient technical support, shortages of specialist lecturers and fragmented institutional planning at the same time, thus creating a capacity trap. Even when an institution acquires an AI platform, lecturers may not know how to integrate it into teacher education. Students may lack stable access outside campus. Curriculum rules may not recognize AI-based practice. Data-governance policies may be absent.

Technology then remains underused, inconsistently applied or dependent on short-term pilot funding. The study's AI-Capacity-Content Interaction Model captures this problem clearly. Effective use emerges only when three elements overlap: capable teachers, appropriate AI tools and a supportive institutional and curricular environment. Investment in one element without the others produces limited value.

For rural South African universities, this is especially important. The country has strong policy commitments to inclusive education, but implementation remains uneven, particularly where universities and placement schools lack resources.

When future teachers receive mostly theoretical preparation and limited exposure to assistive technologies or inclusive classroom practice, AI could help narrow the gap. However, it can do so only when infrastructure, lecturer development and curriculum reform advance together.

The same logic applies across the Global South. Importing systems developed for high-income universities may fail where bandwidth is limited, subscription costs are high, local-language support is weak or assumptions about class size and learner data do not match reality.

The Real Investment Case Is Institutional, Not Technological

The study points to a major policy error: treating AI adoption as a procurement exercise. Universities may be tempted to measure progress by the number of licences purchased, devices distributed or platforms launched. But these indicators say little about whether teachers are becoming more capable or learners are receiving better support.

A stronger investment strategy would begin with clearly defined educational problems. Universities should identify where pre-service teachers struggle most, such as differentiated lesson planning, inclusive assessment, reflective practice or exposure to complex disability-related scenarios, and then select tools suited to those needs.

Short-term priorities include connectivity audits, lecturer training, accessible devices and low-bandwidth or offline-capable applications. Teacher-education programmes should also cover verification of AI outputs, data protection, algorithmic bias and the responsible use of learner information.

Longer-term policy must connect AI adoption with accreditation, teacher standards, disability inclusion and higher-education funding. Governments and development partners should support shared infrastructure, regional training networks and open educational resources rather than isolated technology projects that disappear when grants end.

Businesses also have a role. The strongest opportunity is not simply selling generic AI platforms to universities. It is developing affordable, multilingual and accessible tools for resource-constrained settings. That includes systems designed for intermittent connectivity and applications aligned with local curricula and the needs of learners with disabilities.

Civil society and disability organizations should be involved in design and evaluation. Without their participation, AI tools may reflect assumptions that overlook the lived experience of learners and teachers.

The Global South Must Prevent AI from Becoming a New Divide

AI could help universities overcome shortages of specialist expertise and expand access to inclusive teaching resources. It could also create a new divide between institutions that can afford sophisticated systems and those that cannot. This risk is particularly serious because inequality in teacher preparation has a multiplier effect. If rural universities cannot prepare teachers to use emerging tools, the disadvantage will extend into the schools and communities those graduates serve. The outcome would not be a temporary technology gap, but a widening difference in the quality of education available to learners with disabilities.

The findings connect directly to Sustainable Development Goal 4 on inclusive and equitable education and SDG 10 on reducing inequality. They also raise questions about data sovereignty and technological dependence. Universities relying on external platforms may have limited control over learner data, pricing, language support and future access.

The evidence base, however, remains incomplete. The study is a scoping review, not an impact evaluation. It maps themes across diverse studies but does not establish how much AI improves teaching competence or learner outcomes. It also did not undertake a formal quality appraisal of the included studies, while the underlying research varies widely in method and context.

These limitations mean the findings should guide careful experimentation rather than justify rapid, large-scale deployment. The next phase of research should move beyond perceptions and pilot demonstrations. Rural universities need longitudinal studies comparing AI-supported and conventional teacher preparation, with evidence on teaching performance, accessibility, costs and classroom outcomes. Researchers should also test local-language tools, offline systems and disability-specific applications.