What Should We Be Thinking About?
I remember spending time in a classroom outside Lilongwe, Malawi, in 2022. We were there to research potential learning resources. A colleague asked if the school wanted e-books. On the surface, it made sense ; but the school had only one computer in the admin office and almost no reliable internet. Meanwhile, every teacher had a mobile phone and used WhatsApp daily. They didn’t need e-books; they needed resources that could be delivered via mobile networks. This experience illustrated the risks of assuming “cutting-edge” solutions will work in contexts that don’t share the same infrastructure.
With the rapid rollout of generative AI worldwide, we must avoid repeating this mistake. If we offer Western-based chatbots to African schools, we risk missing the mark on local realities and losing an opportunity to innovate and truly support teachers and students.
Immense Potential, Diverse Challenges
I’m a big fan of Generative AI. It offers promising ways to address some of the biggest educational challenges ; especially in countries where qualified teachers are in short supply or need additional support. For instance, schools in Zambia, Kenya, and Cameroon often grapple with similar obstacles: large class sizes, limited resources, and teachers juggling multiple subjects. An AI-powered teaching assistant could help plan lessons, create assessments, and offer feedback at scale.
Yet, I’ve also seen firsthand how technological assumptions can go wrong. It’s not enough to deploy AI that’s primarily trained on Western knowledge and content; we need local solutions that align with distinct linguistic, academic, cultural, and technological needs. After all, education is not just about imparting facts ; it’s also about reflecting the values that society deems vital.
Risks of a One-Size-Fits-All AI
- Cultural Bias and Technological Colonialism
If models are trained mostly on Western data, they risk reinforcing Western norms as the default. This can perpetuate “technological colonialism,” where AI solutions inadvertently erode local traditions or overlook indigenous knowledge that has not yet been digitized. - Marginalization of Local Knowledge
Many regions rely on oral traditions and community expertise that don’t appear in standard training datasets. By ignoring these knowledge systems, AI might fail to represent crucial parts of local culture and heritage. - Language Exclusion
In countries like South Africa, with 11 official languages, reliance on a single AI that speaks only English or French (common across Africa) could exclude learners from meaningful engagement in their mother tongue. Early experiences with AI chatbots, like ChatGPT’s limited African language capabilities, highlight this issue.
A Call for Context-Specific AI
As Professor Tshilidzi Marwala noted in a recent UNESCO webinar, locally specific AI models are necessary for each region’s unique needs. Policymakers, local experts, and AI developers must coordinate to ensure that:
- Language Diversity is Supported
- AI virtual teaching assistants should handle multiple local languages. While some large language models already respond in Zulu, more investment is needed to ensure all official languages ; including minority languages ; are included.
- Curricula Are Aligned
- AI must be trained on locally approved materials. In South Africa, that means grounding the AI in the CAPS curriculum. AI can deliver contextually accurate instruction using approved textbooks and other vetted resources.
- Pedagogical Approaches
- Teaching is not just about content; it’s about methods and values. AI assistants should mirror local pedagogical techniques such as group work, inquiry-based learning, or problem-solving approaches outlined in the national curriculum.
- Values Are Embedded
- Accessibility Is Prioritized
- Many educators and learners rely on mobile devices and 3G networks. A genuinely inclusive AI solution should work on WhatsApp or low-bandwidth channels. Tools like Maski, an AI teacher assistant via WhatsApp, show how a carefully designed platform can make a real difference in under-resourced areas.
Envisioning an Equitable AI Future
Generative AI has the potential to revolutionize education in the Global South by supplementing teacher capacity, personalizing learning, and expanding access. But that potential can only be realized if we confront the challenges head-on:
- How do we invest in local AI research and development so Western companies do not just drive it?
- How do we digitize and include local knowledge bases, oral traditions, and indigenous languages?
- How do we empower communities to shape AI, rather than merely consume it?
We have the power to shape education through responsible and culturally responsive AI. But designing tools that truly empower learners requires collaboration among governments, educators, tech companies, and local communities.
Let’s seize this moment to shape the future of education in a way that celebrates diversity, respects local context, and fosters equitable access to knowledge. The road ahead may be challenging, but the reward ; an education system that serves every learner ; is well worth the journey.