Generative AI and Education Systems in the Global South — What Should We Be Thinking About?

I remember spending some time in a classroom in a school outside Lilongwe, Malawi, in 2022. We were visiting to research developing new resources for schools in the country. A colleague asked me to find out if they wanted e-books. The school had one computer in the admin office and no reliable internet connection. However, all the teachers had mobile phones and used WhatsApp. The question needed to be framed correctly. The teachers did not need e-books but could benefit from resources and support delivered via mobile networks (not an attractive business model for publishers!). With the rollout of generative AI globally, a similarly wrong framing could occur based on the assumption that African schools could use Western chatbots to improve their education.

Generated by the author and Dall-E

I’m a big fan of Generative AI and think it has immense global potential for education systems, particularly in less-developed countries, where I’ve spent most of my professional life. These countries often don’t have enough qualified teachers, or the teachers they have need additional support. Schools in countries as diverse as Zambia, Kenya and Cameroon share similar issues in this regard. How valuable would a virtual AI teaching assistant be in helping with lesson planning, assessment, or feedback for learners in these schools? These are a few projects I’ve worked on recently, and I think they have great potential. However, this is again an assumption that most school systems in the Global South work similarly and need similar things to those in the North. This is not the case, as I have seen with the adoption of digital learning technology across African schools, which focuses on mobile-first and offline access. A discussion I had last week got me thinking about what it might mean for these schools and teachers to use an off-the-shelf AI solution developed by Western organisations and trained primarily on Western knowledge and content. After all, education is about teaching knowledge and skills and the values a country feels are essential.

For this technology to fulfil its potential in education across the developing world, it must be consciously adapted to each society’s linguistic, academic, cultural, and technological ecosystem. Current models are unlikely to meet these needs fully. There are potential risks related to cultural bias where norms are skewed towards Western countries’ standards; the marginalisation of local knowledge, particularly in regions still reliant on oral transfers of knowledge that have yet to be digitised; and a type of insidious technological colonialism that could influence the minds of teachers and students, even if unconscious.

As Prof. Tshilidzi Marwala stated in a recent UNESCO webinar, education systems need locally specific AI models tailored to their unique context and requirements. For example, as an education ministry develops its curriculum, it should build its framework for incorporating AI. This requires significant investment in research and development led by local education experts and government stakeholders.

The following highlighted what the South African education ministry, keen to incorporate AI tools in their schools, might need a model to do. In South Africa, the schools are predominantly government-run, following the national curriculum called CAPS. Part of education in South Africa is to instil social cohesion in a diverse society with a fractured history. Therefore, educational resources must celebrate this diversity and align closely with the country’s constitution. This means that an AI model for use in South African schools needs:

  • Language Diversity: In countries like South Africa, which has multiple official languages, AI virtual teaching assistants must be able to engage with learners in their mother tongues. Models should be trained on diverse linguistic data to communicate effectively in all official languages. African language models must be developed further, but chatbots like ChatGPT can respond to popular African languages such as Zulu.
  • Curricular Alignment: Virtual teachers should be trained on materials approved for the specific national curriculum (e.g. CAPS in South Africa). This ensures that AI imparts knowledge that is aligned with educational standards and local perspectives. One way to do this is to use approved textbooks from the national catalogue, which have been written for the curriculum and are for use in South African classrooms. This could raise issues around copyright and fair use.
  • Pedagogical Approach: AI teachers must mirror local educators’ instructional styles and methods. In South Africa, this means promoting critical thinking, problem-solving and knowledge integration as outlined in the CAPS principles section of the curriculum.
  • Instilling Values: Most importantly, AI in education must deliberately be designed to nurture democratic values, human rights, and social-emotional competencies essential for engaged citizenship in the country’s unique historical and cultural context. Fundamental values to embed in South Africa include social justice, equality, ubuntu, accountability, empathy, diversity and more.
  • Accessibility: With mobile-first populations, AI learning solutions should be designed for ubiquitous platforms like WhatsApp and be lightweight enough to function well on essential smartphones and 3G networks. In South Africa, initiatives like Maski (an AI teacher assistant available via WhatsApp) illustrate how thoughtfully designed AI tools could dramatically expand educational opportunities, even in resource-constrained contexts.

Generative AI holds immense potential to revolutionise education systems in the Global South, offering opportunities to bridge educational gaps, support teachers, and provide personalised learning experiences. However, to fully realise this potential, we must proactively address the challenges and risks of implementing AI in diverse cultural and linguistic contexts.

As we move forward, we must ask ourselves: How can we ensure that the development and deployment of AI in education systems in the Global South is inclusive, equitable, and respectful of local contexts? It is a complex challenge that requires the collective efforts of governments, educators, AI developers, and communities.

We have a unique opportunity to shape the future of education through the responsible and culturally responsive integration of AI technologies. Let us seize this moment to create AI-powered education systems that empower learners, support teachers, and contribute to the sustainable development of communities across the Global South. The journey ahead may be challenging, but the potential rewards — a more equitable and accessible education for all — are well worth the effort.

Leave a Reply