How AI can transform Africa's learning crisis into a development opportunity

Consider these sobering stats: 89 per cent of children in sub-Saharan Africa cannot read a simple text by the age of 10 (referred to as "learning poverty"), and by 2050, two in five children will be in Africa. Today, more than 30 million young people are not in education, training or employment. Creating and scaling up solutions to help more children learn has never been more urgent.

AI promises much here. Yet its deployment in African education and work places (formal and informal) requires careful consideration. Recent research, examining how humans interact with generative AI systems, suggests AI’s impact may prove to be a double-edged sword.

AI as productivity enabler or cognitive risk

From Offloading to Engagement by Prof. Dr. Michael Gerlich looked at how 150 participants in the UK, Germany and Switzerland (covering secondary school students to working professionals) engage with generative AI. Four conditions were tested: human-only, human + AI (unguided), human + AI (guided with structured prompting), and AI-only benchmarks.

Results showed that "unguided AI use fosters cognitive offloading ... whereas structured prompting significantly reduces offloading and enhances both critical reasoning and reflective engagement." Structured prompting is an approach that requires users to pause, reflect, and test their own assumptions as part of self-regulated learning.

In this study, the human + AI (guided) condition was designed to promote deliberate engagement with an AI system while completing a task. In short, the AI served not as an answer machine but as a thinking partner. The study concludes that the difference is key to successful AI adoption in society and business.

The need for ecosystem support

Zooming out, lessons from years of EdTech interventions by UNICEF have shown that success depends not just on good software, but on teachers. They provide a level of human support and scaffolding critical for digital learning. Quality content, equal access, accessibility, and implementation research to measure impact, are other critical components of the learning ecosystem.

If these are the preconditions for successful use of AI to learn and to be productive at work, they present major challenges for Africa. These are not the default conditions in the classroom. As we've seen, there is widespread learning poverty amongst students. AI literacy is not commonplace in schools.

Recent research with children on AI by the Digital Futures for Children centre (DFC) and Mtoto News, revealed that AI literacy among the participants in Kenya was primarily self-taught.

"We should have a Kid AI class where we learn safely", noted Pinkie (13), as study participants said AI was the future.

Teachers often have insufficient subject matter expertise and low motivation, which stem from inadequate support and poor school and digital infrastructure. In such contexts, there is a temptation to outsource thinking to AI and use it to do, versus support, work.

The promise of AI in Africa

The stakes are high to getting AI right in Africa. Genesis Analytics estimates that by 2030, AI could inject $2.9 trillion into the African economy, equating to a 3 per cent annual increase in GDP. This would potentially lift 11 million Africans out of poverty and create half a million jobs across the continent each year. Some young workers in Africa are already integrating AI into their daily routines, using tools like ChatGPT, Copilot, and in-house chatbots to enhance productivity, creativity, and efficiency. A forthcoming UNICEF paper points to how AI is supporting teachers and learners in promising practices.

Investment in skills is key to unlocking this collective potential. It is not surprising that the African Union Continental AI Strategy has a focus on children and youth that includes supporting AI skills, AI in education, AI for economic opportunities and employment, and AI research and innovation. The Africa Declaration on AI, endorsed by most African countries earlier this year, equally pays attention to skills and capacity. "Continent-wide AI education initiatives with curriculum development for youth at all levels" is a key commitment.

The private sector is also getting involved: Microsoft aims to develop the digital and AI skills of a million South Africans by 2026, while the University of Lagos is the first African institution to launch an OpenAI Academy for capacity building. Mastercard Foundation invests in digital skills, entrepreneurship and workforce development in Africa. Generation Unlimited, partnering with companies and governments, provides skilling programmes across the continent.

Moving forward with intention

These efforts are welcomed, but it is no secret that a lot more needs to be done. The Gerlich study highlighted the "dual nature of AI as both a productivity enabler and a cognitive risk". The latter is a risk Africa cannot afford. As African nations seek to embed AI in education and the workplace, the distinction between augmenting and replacing human cognition becomes crucial.

Simply providing access to AI tools (it is common in digital learning interventions to lead with technology provision) could prove counterproductive. With One Laptop per Child, the fallouts included wasted expenditure and poor learning outcomes. Unguided use of AI systems may actively deskill students and teachers.

More action, more research

A holistic approach is needed: Training teachers to effectively support AI usage of their students, and to use the tools for their own tasks such as lesson planning, administration, and certain types of assessment. African students need to be digitally and AI literate.

AI systems need to be designed for local contexts, local languages and local curricula, to suit the needs of teachers and workers. With such systems, that have scaffolding mechanisms like structured prompting, coupled with supportive teachers and skilled learners, the potential is enormous.

But, as Maria Barron of the World Bank suggests, more research is needed to understand how to design AI systems to support learning and minimize cognitive offloading. We need to implement and evaluate how to best empower teachers to use AI systems well, and support their classes to do the same. This can only be achieved by engaging children, youth and teachers in African countries to ensure solutions are local and inclusive.

At UNICEF we've started a project about the future of skills in an AI world, with a focus on Africa. Watch this space.