AI in African Higher Education: Innovation’s Promise, Inequality’s Warning

Artificial intelligence is already moving into classrooms, admissions systems, grading tools, learning platforms, administrative workflows, and student-support services. For universities facing rising enrolments, stretched faculty, uneven resources, and pressure to modernize, AI promises to deliver faster feedback, personalized learning, better resource allocation, and more efficient institutions.

However, a new study on South Africa and Kenya argues that this promise carries a sharper question: can AI make universities better without making them less equal?

The paper, "Navigating AI in Higher Education: Balancing Efficiency, Equity, and Autonomy in South Africa and Kenya," by Mahlatse Given Sevhake and Costa Hofisi of North-West University, examines how AI adoption is reshaping higher education in two African countries with different governance systems and infrastructure challenges. Published in AI in Education, the study uses qualitative document analysis of policy frameworks, scholarly literature, and institutional reports to assess the tensions between efficiency, equity, and student autonomy.

The Efficiency Promise Is Real, but Not Neutral

Intelligent tutoring systems, automated grading, predictive analytics, and digital feedback tools can help institutions respond faster to students and reduce pressure on educators. The study notes that AI tools can support personalized learning, streamline administration, and improve educational quality.

In higher education systems, faculty workloads are rising and institutions are expected to serve large, diverse student populations with limited resources. AI can identify struggling students earlier, customize learning pathways, speed up routine tasks, and free educators to focus on higher-value academic work.

The study warns that efficiency is not automatically justice. When universities optimize for speed, automation, and prediction, they may weaken the deeper purpose of education: critical thinking, creativity, dialogue, and independent judgment. The paper finds that AI adoption can improve efficiency, but it can also reduce metacognitive reflection and critical engagement when learners become overly dependent on algorithm-generated feedback.

Old Inequalities Can Become New Algorithms

According to the study, AI does not enter a blank slate. It enters education systems already shaped by history, geography, language, income, infrastructure, and institutional privilege.

In South Africa, AI adoption is filtered through the legacy of apartheid-era inequality and uneven access across historically advantaged and disadvantaged universities. The study reports that 92 percent of students in historically advantaged universities experience stable internet access, compared with only 58 percent in disadvantaged universities. It also warns that AI tools used in admissions, placement, or assessment may disadvantage students from poorer schools and non-English-speaking backgrounds.

In Kenya, the divide is more closely tied to infrastructure and socioeconomic inequality. While more than 95 percent of the population reportedly has access to mobile networks, only 35 percent has internet access overall. The gap is stark: 56.6 percent of urban populations have internet access, compared with 25 percent in rural areas.

These numbers are crucial because AI-enabled education depends on connectivity, devices, data access, and institutional readiness. Students without stable internet cannot benefit equally from adaptive learning tools, AI tutors, digital feedback systems, or online student-support platforms. In that context, AI can widen the distance between students who are already connected and those still struggling for basic digital access.

The risk is not simply that some students miss out. It is that exclusion becomes embedded into automated systems. If AI tools are trained on historical data that reflect unequal schooling, language privilege, or urban advantage, they may reproduce those patterns under the appearance of technical neutrality.

Bias Is Not Just a Technical Bug

The study argues that algorithmic bias in higher education should not be treated as a narrow coding problem. It is a governance problem. In South Africa, historical enrolment data may recreate apartheid-era inequalities. Linguistic bias may work against students whose first language is not English. In Kenya, scholarship allocation algorithms may favor urban and economically advantaged students because those students are more likely to have the academic and extracurricular profiles rewarded by automated systems.

Universities cannot simply buy or deploy AI tools and assume fairness will follow. They need transparency rules, bias audits, appeal mechanisms, multilingual design, and student protections. Students should know when AI is influencing decisions that affect their learning, scholarships, assessments, or progression. Educators should be able to question algorithmic outputs rather than defer to them.

The study highlights that South Africa and Kenya both have regulatory foundations, including South Africa's Protection of Personal Information Act and Kenya's Data Protection Act, but it also points to gaps around fairness, bias, and context-specific accountability. The key issue is not whether AI can be useful in universities; it's whether institutions have the legal, ethical, and pedagogical safeguards to ensure that AI supports inclusion instead of automating privilege.

The Future of Learning Must Keep Students in Control

AI can personalize learning, but it can also guide students into predetermined pathways, nudge behavior through predictive analytics, and shift decision-making power from learners and educators to platforms. The authors warn that AI-mediated learning environments can move the "locus of control" away from both students and teachers toward algorithms. They call for a human-centered approach in which learners retain the right to doubt, question, and disregard AI recommendations.

Higher education is not only about delivering information. It is about forming people who can think independently, challenge assumptions, participate in civic life, and make judgments in uncertain situations. If AI turns students into passive recipients of machine-curated pathways, universities may gain efficiency while weakening agency.

AI in higher education must be treated as part of the broader development agenda. It connects to SDG 4 on quality education, SDG 9 on digital infrastructure and innovation, SDG 10 on reduced inequalities, and SDG 16 on accountable institutions. For the Global South, the stakes are especially high because digital divides and institutional inequalities are already deep.

The study is based on document analysis rather than new field interviews, surveys, or system testing, and no new primary data were generated. Its strength lies in synthesis and policy interpretation, not direct measurement of AI outcomes across campuses. Even so, its findings are crucial. AI could help African universities expand access, personalize support, and manage growing demand, but without inclusive infrastructure, ethical literacy, participatory governance, and strong accountability, it could also make old inequalities faster, quieter, and harder to challenge.