Learning is in crisis. Even before COVID-19, 53 percent of children in low- and middle-income countries (LMICs) were unable to read and understand a simple text by the age of 10. The pandemic is expected to push that figure up to 70 percent.

The past decade has generated major advances in technology and experimentation with digital learning solutions that enable a new kind of experience – by tailoring learning to the needs of the individual; what we are calling digital personalized learning.

Digital personalized learning has shown promise in LMICs in closing education gaps for lower-attaining students by allowing them to learn at their own pace and to their own proficiency, positioning it as a potential tool to address learning gaps as the worst of the pandemic recedes.

Taking stock of digital personalized learning in the developing world

So we at UNICEF decided to take stock of digital personalized learning in LMICs to understand how these solutions work and to provide evidence-based insights to inform their design and implementation. Our first report in the series examines trends in design and implementation, based on a review of 40 personalized learning EdTech products deployed in LMICs.

Market penetration

The market for these products in LMICs is growing and is partly being driven by local entrepreneurs – with 70 percent of products headquartered out of an LMIC. But market penetration remains uneven: in particular, it is weak in francophone central and west Africa, and in low-income countries.

Content and pedagogy

While the products we reviewed target both primary and secondary levels, they tend to focus on few subjects – mainly math and language/literacy. Evaluations, too, have focused disproportionately on impacts in these subjects, while the evidence on impacts in other subjects, such as the sciences and 21st century skills, remains thin.

They were also more than twice as likely to target supplemental learning rather than core learning – whether in school or at home. And while content is offered in multiple local languages in some regions, local language options are very limited among products deployed in sub-Saharan Africa.

Contrary to the perception that these solutions bypass the teacher, most are designed to engage teachers in student learning, though this is mainly limited to monitoring student performance. Yet very few products translate this data into recommendations that teachers can act on to support student learning, such as follow-up interventions. Meanwhile, only a minority of products share data with parents on their child’s progress.

Future design and implementation should aim for more data transparency with key stakeholders like teachers, parents and school leaders, enabling them to engage in meaningful ways and to build their capacity to support children in their learning journey.

Product design and technology

Some 60 percent of products leverage artificial intelligence to dynamically adapt learning paths based primarily on student performance on learning tasks. However, more research is necessary to understand the effectiveness, accuracy and bias of different algorithmic approaches.

These products also demand levels of hardware and connectivity that can be tough for many children in LMICs to access – very few can be used on feature phones and fewer than half offer any kind of offline access.

We also found gaps in data protection policies that must be addressed to ensure the safety of students and their information.

Inclusiveness

To some extent, the products we reviewed cater to diverse populations. Unfortunately, students who are already disadvantaged remain underserved, including learners with disabilities, out-of-school children, ethnic minorities and displaced children. Yet these groups need remedial learning support the most, so it is important to consider how to better support these marginalized learners.

Results measurement

Fewer than half of these solutions have evaluations available to the public. And the evaluations that are available rarely conducted cost-benefit analysis or looked at impacts at-scale. Comparing one product to another is difficult, so there is a need for stronger research-product partnerships and for standards for comparison between solutions – on factors such as quality, efficacy and ease of use.

The way forward

For years, educators have advocated for the transformative potential of personalized learning. To realize this potential, and to ensure these digital solutions contribute to mitigating the learning crisis, much more needs to be done.

We want to see more support for equitable access and localization; more attention to connectivity constraints; a robust research agenda around measuring learning impacts, cost effectiveness, and the relative effectiveness of different design and implementation options; stronger government ownership; and more attention to alignment with industry standards on issues like interoperability and data protection, among others. This research aims to build out guidance to help us get there faster.