Artificial intelligence is becoming an integral part of education, already making its way into classrooms, admissions systems, grading platforms, learning tools, administrative functions, and student support services. For universities grappling with rising enrollment numbers, limited faculty, disparate resource allocation, and the imperative to modernize, AI presents the opportunity for swifter feedback, personalized learning experiences, improved resource management, and enhanced organizational efficiency.
However, recent research focusing on South Africa and Kenya raises a critical question: Can AI contribute to greater equality within these institutions rather than exacerbate existing disparities?
The paper “Navigating AI in Higher Education: Balancing Efficiency, Equity, and Autonomy in South Africa and Kenya,” authored by Mahalatse Given Sebhake and Kosta Hoffisi from North West University, explores how AI’s implementation is transforming higher education in these two African nations, which face different governance systems and infrastructural hurdles. This study, published in AI in Education, employs qualitative document analysis of policy frameworks, academic literature, and institutional reports to examine the balance between efficiency, equity, and student autonomy.
Efficiency Offers Tangible Benefits, but Lacks Neutrality
AI-driven tools such as intelligent tutoring systems, automated grading, predictive analytics, and digital feedback mechanisms are enabling institutions to respond more rapidly to student needs while alleviating pressure on educators. The findings indicate that these technologies can enhance personalized learning, streamline administrative processes, and improve educational quality.
As faculty workloads continue to grow and educational institutions are tasked with serving an increasingly diverse student population under resource constraints, AI can identify at-risk students early, tailor learning pathways, and expedite routine tasks, allowing educators to concentrate on higher-level academic pursuits. Yet, the study warns that enhanced efficiency does not inherently equate to fairness. As universities prioritize speed, automation, and predictive insights, they risk undermining the essential educational objectives of fostering critical thinking, creativity, dialogue, and independent judgment. The research suggests that while AI may boost efficiency, it could inadvertently stifle metacognitive reflection and critical engagement, as students may become overly reliant on algorithm-driven feedback.
Existing Inequities May Resurface as Algorithmic Biases
Research indicates that AI cannot be seen as a neutral entity; it reflects existing educational systems already influenced by historical, geographical, linguistic, income-related, infrastructural, and institutional factors. In South Africa, the integration of AI is influenced by the enduring legacy of apartheid, which has resulted in unequal access between historically privileged and underprivileged institutions. The study reveals that 92 percent of students at historically advantaged colleges have reliable internet access, compared to only 58 percent at their disadvantaged counterparts. It cautions that AI applications for admissions, assessment, and placement could further disadvantage students from poorer schools and those who are non-English speakers.
In Kenya, the digital divide is more closely tied to infrastructural and socioeconomic disparities. While over 95% of the population reportedly has mobile network access, only 35% have internet access. The contrast is stark, with urban residents enjoying a 56.6% internet access rate compared to just 25% in rural areas. These statistics are crucial, as AI-enhanced education relies heavily on connectivity, device accessibility, data availability, and institutional preparedness. Without reliable internet, students miss out on the benefits of adaptive learning tools, AI tutors, and online support systems. Consequently, AI could widen the gap between digitally connected students and those who struggle with fundamental internet access.
The concern extends beyond mere absence; exclusion can be embedded within automated systems. When AI is trained on historical data reflective of inequitable schooling, linguistic advantages, and urban bias, it risks perpetuating these disparities under the guise of technological impartiality.
Algorithmic Bias Is a Governance Challenge
The study contends that algorithmic bias in higher education must be viewed as a broad governance issue rather than a mere technical flaw. In South Africa, historical registration records could potentially reinstate apartheid-related inequalities. Language biases may disadvantage students whose first language is not English, while in Kenya, scholarship distribution algorithms may inadvertently favor economically privileged urban students, as these individuals tend to possess academic and extracurricular credentials more easily recognized by automated systems.
Universities cannot simply implement AI tools and assume they will ensure equity. To truly harness the potential of AI, institutions require transparency measures, bias audits, appeal processes, multilingual options, and robust student protections. Students must understand how AI influences decisions impacting their education, scholarships, assessments, and academic progress, while educators must remain vigilant and critical of algorithmic outputs.
Although both South Africa and Kenya possess regulatory frameworks, such as South Africa’s Personal Data Protection Act and Kenya’s Data Protection Act, there remain significant gaps concerning fairness, bias, and contextual accountability. Rather than questioning if AI can assist universities, it is more pertinent to ask if institutions have established the necessary legal, ethical, and educational safeguards to ensure that AI encourages inclusivity rather than perpetuating privilege.
Student Agency Must Be Central to Future Learning
While AI holds the promise of personalizing education, it may also inadvertently steer students down predetermined paths, utilizing predictive analytics to shift decision-making power away from learners and educators. The authors caution that AI-mediated environments can reallocate control from both students and teachers to algorithms. They advocate for a human-centered approach where students maintain the right to question, critique, and even disregard AI recommendations.
Higher education transcends mere information delivery; it aims to cultivate independent thinkers who challenge assumptions, engage in civic life, and navigate uncertainty. Should AI reduce students to passive recipients of machine-curated pathways, universities stand to gain efficiency at the expense of student agency.
AI in higher education should be regarded as part of a broader development agenda, contributing to Sustainable Development Goals (SDGs) related to quality education, digital infrastructure and innovation, reducing inequalities, and fostering accountable institutions. This is particularly relevant in the Global South, where pronounced digital divides and systemic inequalities are already prevalent.
The study’s foundation lies in document analysis, not in new fieldwork, surveys, or system evaluations, resulting in no new primary data. Its strength emerges from policy interpretation and integration rather than direct measurement of AI’s impacts across campuses. Nonetheless, the findings underscore a crucial point: while AI could enable African universities to broaden access, personalize support, and cope with increasing demand, it also has the potential to deepen existing inequalities—subtly, quickly, and without the necessary infrastructure, ethical understanding, participatory governance, and robust accountability in place.
