Can AI transform learning for the world's most marginalized children?

Recent advances in artificial intelligence (AI) have sparked a growing interest in the potential impact it can have on education. Like any technology, AI presents opportunities and challenges. Unfortunately, past advances in technology have often favoured the privileged, exacerbating inequalities in education and other sectors. Will AI follow the same path? Or, does it hold the potential to reduce education inequalities and transform learning for the most marginalized?

When considering the broader impact of AI on education, there is no doubt that it will be transformative — it is only a matter of time. Many education and technology leaders have weighed in on the potential – and pitfalls – of AI for education. Bill Gates, for example, believes that soon AI tools will be able to help children learn how to read and write. On the other side of the spectrum, many critics have decried the use of AI, pointing out that it helps students cheat and can deliver biased results, including gender bias, and is prone to making errors and presenting them as fact.

We need to be realistic though about the global impact. While almost all countries rolled out digital learning platforms during the COVID-19 pandemic, UNICEF’s Pulse Check on Digital Learning research revealed that one-third of those platforms are no longer maintained or updated and the vast majority lack interactive content. Transforming these platforms from repositories of learning content into interactive platforms – let alone integrating AI-powered solutions – requires significant effort and investment.

Some governments are making that commitment to digitally transform their education systems at a rapid pace, including the piloting of AI solutions. Through the Gateways to Public Digital Learning initiative, UNICEF together with UNESCO is collaborating with these pioneering countries in Africa, Asia, Latin America and other parts of the world to share best practices and solutions. The aim is to showcase the potential of innovative digital learning to transform education and to turn digital learning into a public good – freely accessible to all.

For UNICEF, a key priority is to ensure that digital learning, including AI technologies, supports children in emergencies and children outside of the formal school system. Globally, 222 million children are affected by emergencies and protracted crises; 244 million children are out of school and 1 billion children — nearly half the world’s children — live in countries classified as “extremely high-risk” to the impacts of climate change. Schools in these countries close frequently due to natural disasters, such as floods, hurricanes and storms.

Hundreds of millions of children lack access to tablets, computers or reliable Internet connectivity at home, posing challenges to home learning when schools close. To reach the most marginalized at scale, we need to consider how solutions leveraging AI can run on shared, low-cost smartphones and function offline. We also require more and better digital learning solutions to help children acquire foundational literacy. Nearly two-thirds of 10-year-olds are estimated to be unable to read and understand a simple text. AI solutions, such as ChatGPT, are of little use for those who have yet to learn how to read and write, unless they embed multilingual speech recognition.

Regardless of advances in AI tutoring systems and other technologies that support independent learning, teachers should remain central to the learning process. This is also true for home-based learning during school closures. Our research revealed that the frequency of contact with teachers significantly correlated with self-reported learning during COVID-19. So, we see that AI has huge potential to support teachers, not replace them.

In many countries, less than half of schools have devices for learning. For schools with devices, they are often shared amongst many students, or only accessible to teachers. Nevertheless, the potential is still huge. AI solutions can enable personalized learning experiences even on shared devices – and help transform schools and teaching practices to become more learner-centred.

Lack of digital skills is also a key barrier to using technology. A significant gender digital divide exists in terms of access to technology, as well as digital skills — based on UNICEF’s latest research across 32 countries and territories, for every 100 male youth with digital skills, there are only 65 female youth. AI can support learners with low digital literacy by incorporating features such as speech recognition and natural language processing. It can also aid in digital skills acquisition through adaptive systems that automatically adjust to the learner’s literacy – as well as digital literacy - levels. However, many of these technological advancements are behind paywalls and intended for high-end devices. Significant investments and efforts are needed to make them freely accessible to all.

Another critical area of focus for UNICEF is accessible digital learning for children with disabilities. Out of the 240 million children with disabilities globally, the majority lack access to inclusive technologies, accessible learning materials and other essential education support.

AI can make a substantial difference in this area as well. Leveraging voice recognition across different languages and dialects, text-to-sign language automation and ChatGPT’s ‘vision’ to interpret images and diagrams could all be useful. UNICEF has been working for several years with governments on developing accessible digital textbooks, but it has thus far been a costly and time-consuming process. Additionally, personalized learning that leverages AI to adapt to learner abilities and interests can benefit children with disabilities. UNICEF is now working with partners on a series of consultations with AI and disability specialists, researchers and developers to ideate on scalable AI-based solutions for accessible learning.

To address the question posed in the title of this blog: can AI transform learning for the most marginalized? the answer is a resounding yes. Time will tell, however, whether AI will reduce or further widen inequalities. Significant investments and a concerted focus on designing and developing solutions for the most marginalized are essential. We firmly believe that this transformation is possible through collaboration with governments, civil society, the private sector and other partners. Together, we can make it happen.