A study involving researchers from the Kwame Nkrumah University of Science and Technology (KNUST) has warned that introducing artificial intelligence into African universities without changes to teaching, institutional policy and access to resources could deepen existing inequalities. The research, published in Frontiers in Education and reported by KNUST on October 22, examines how people across five sub-Saharan African countries experience AI education and how prepared institutions are to support it.
The findings are relevant as universities, lecturers and students increasingly encounter AI tools for research, writing, assessment, data analysis and learning support. The researchers argue that access to a tool is not the same as the ability to use it meaningfully. Students need devices, connectivity, training and clear guidance, while lecturers and institutions need time, resources and policies that explain how AI can be used responsibly.
The study was led by Dr Jeffrey N. A. Aryee and conducted under the Revolutionizing Tertiary Education for Africa’s Thriving AI Economy and Workforce (RetAIn) project. Its message is not that universities should reject AI. Rather, the authors say the education system must develop alongside the technology so that adoption does not advantage only those who already have better access and institutional support.
What the research examined
The researchers combined focus group discussions involving 77 participants in Ghana, Rwanda and Namibia with a survey of 156 respondents across those three countries as well as Kenya and Zambia. KNUST said the study explored awareness of AI, participation in AI-related activities, access to learning resources and institutional preparedness. The authors also examined how different national and university settings approach AI integration.
The research has an important limitation: the survey was exploratory and based on a relatively small, non-random sample. KNUST explicitly cautioned that the findings are not nationally representative. The results therefore should not be read as a precise estimate of how every university student or lecturer in Ghana or the wider region uses AI. They are evidence of patterns and concerns that institutions can investigate further.
That caveat does not make the findings irrelevant. Exploratory research can identify barriers that deserve attention, particularly when the experiences of students and staff are not evenly distributed. The study offers a starting point for universities to ask whether their own policies, facilities and teaching practices are keeping pace with the tools that students are already encountering.
Cost remains a major barrier
About half of survey respondents identified financial obstacles to AI education. KNUST listed tuition fees, certification costs, internet data charges, transport and digital devices among the challenges. More than one-third reported infrastructure limitations, including unreliable or expensive internet connectivity. These barriers can affect whether a student is able to practise with an AI tool, complete an online course or access learning materials outside campus.
For students who have a laptop, reliable data and a quiet place to study, AI-supported learning may be relatively easy to explore. For someone who depends on a shared phone, has limited data or must travel to reach a computer laboratory, the same opportunity may be much harder to use. If assignments and course expectations increasingly assume that students have digital access, the gap can become an academic disadvantage rather than simply a technology gap.
The study also found that students and part-time staff faced greater constraints in accessing learning resources than full-time and contract staff. That distinction matters because part-time learners and staff may have less time on campus or fewer opportunities to attend training. A university can announce an AI initiative and still leave some of its community unable to participate in a meaningful way.
Who risks being left behind?
KNUST said the researchers identified women, persons with disabilities, students from rural and low-income communities, and people combining work with part-time studies as groups at risk of exclusion. Those groups are not uniform, and the study does not suggest that every person in them has the same experience. The concern is that existing inequalities in access, time, infrastructure and support can be reproduced when AI becomes part of everyday teaching.
Accessibility is one example. An AI-enabled learning platform may not be useful to a student if it is incompatible with assistive technology, difficult to navigate or dependent on a high-speed connection. Language is another issue. Tools trained primarily on widely represented languages may not serve students equally when learning and research involve local languages or culturally specific knowledge. Universities need to evaluate these conditions rather than assume that a tool is inclusive because it is available online.
Gender and income gaps can also interact with institutional expectations. If students are expected to purchase a certification, subscribe to a service or complete a project using a personal device, those with fewer resources may be disadvantaged. A fair approach would provide accessible alternatives, campus resources and transparent expectations so that grades reflect learning rather than the ability to pay for technology.
Different countries, different implementation challenges
The KNUST summary describes varied approaches across the countries included in the focus groups. Rwanda’s policy-led approach was associated with implementation difficulties at institutional level. Namibia’s universities showed adaptability but faced challenges keeping curricula aligned with industry practice. In Ghana, institutional initiatives existed alongside gaps in coordinated policy implementation and guidance.
These comparisons are not a ranking of national education systems. They show that having a strategy, having an institution ready to implement it and having students able to benefit from it are separate stages. A national framework can set direction, but universities need procedures, training, budgets and support services. An institution may experiment with AI while still lacking a consistent policy on privacy, assessment or the treatment of AI-generated work.
For Ghanaian universities, the practical question is how to coordinate experimentation without leaving students and lecturers to interpret unclear rules on their own. Institutions may need to specify which uses are permitted, when AI assistance should be disclosed, how student data is protected, and how lecturers can design assessments that measure understanding. Policies should be understandable and updated as the technology changes.
Training must accompany the tools
The researchers called for stronger institutional AI policies, investment in digital infrastructure, improved access to learning resources and attention to local languages and underrepresented groups. They also emphasized governance frameworks covering responsible use, data privacy, accountability and ethics. These recommendations point to a broader definition of readiness: the presence of software is only one component of a functioning AI education programme.
Lecturer training is central because teachers shape how tools are used in the classroom. Training can help educators identify appropriate applications, design learning activities that require critical thinking, explain limitations and avoid treating AI output as automatically correct. It can also support assessment methods that reward process, reasoning and the ability to verify information rather than only the final text submitted.
Students need similar guidance. They should learn how to check factual claims, protect personal information, recognize bias, cite sources and understand when AI assistance is inappropriate. These skills are useful even for students who choose not to use AI. Digital literacy is not simply learning how to enter a prompt; it includes evaluating what a system returns and knowing when to seek a human or authoritative source.
AI can support learning, but it cannot replace the education system
AI tools can assist with brainstorming, explanations, language support, coding, data analysis and feedback. Their usefulness depends on the task, the quality of the tool and the user’s ability to evaluate the result. A fluent answer may contain mistakes, omit context or reflect biases in the data used to develop the system. Universities therefore need to treat AI as a resource that requires oversight, not as a substitute for subject knowledge or academic judgment.
The study’s warning is especially important for research and assessment. If students rely on a system to generate an argument without understanding it, they may miss the learning objective. If a lecturer uses automated tools to evaluate student work without checking their limitations, the process may introduce unfairness. Clear rules and human responsibility remain necessary even when AI makes some tasks faster.
At the same time, refusing to engage with AI could leave graduates less prepared for workplaces where these tools are increasingly used. The challenge is to develop a balanced approach that supports experimentation while protecting academic integrity, privacy and equitable access. Universities can create supervised opportunities for students to practise rather than leaving adoption to informal and unequal use.
What universities and policymakers can do
First, institutions can assess access to devices, internet connectivity, laboratories and assistive technologies before introducing AI-dependent assignments. Second, they can provide training for lecturers and students, with practical examples connected to different disciplines. Third, they can publish clear policies that explain acceptable use, disclosure, privacy and accountability. Fourth, they can monitor who participates and who is left out, using feedback to adjust programmes.
Governments and partners can support shared infrastructure, affordable connectivity and research capacity. Universities can work with industry without allowing commercial tools to determine educational priorities. Local-language resources and accessible design should be part of the planning process from the beginning, not treated as optional additions after a system is already in use.
Finally, institutions should evaluate whether AI initiatives improve learning outcomes and access. Useful measures may include student participation, lecturer training, availability of devices and connectivity, accessibility, and the quality of student work. The study does not provide a universal blueprint, but it gives universities a reason to collect evidence about their own settings and make decisions based on actual needs.
Conclusion
The KNUST-linked study presents AI integration as an education and equity challenge as much as a technology opportunity. Uneven access to devices, data, training and institutional guidance can make the benefits of AI easier to reach for some students than others. The researchers caution that their sample is exploratory, but their recommendations are clear: strengthen policies, improve resources, train educators and learners, and protect the groups most likely to face barriers.
For African universities, the goal should be meaningful and responsible use rather than adoption for its own sake. When teaching practices, institutional systems and access develop alongside AI, the technology may support learning and workforce preparation. Without those changes, it risks reinforcing the very inequalities education is meant to reduce.
Sources and related reading
Read KNUST’s report on the study and the university’s profile of Dr Jeffrey N. A. Aryee. Related reading on MyGHPages includes digital skills and employability in Ghana and our report on digital cross-border payments.
More Ghana and Africa technology, education and business coverage: MyGHPages.com.
