Artificial intelligence can extend support, translate explanations and help educators respond faster. It cannot replace the human and physical conditions that make learning possible.

Knowledge ArkSeptember 23, 20268-minute read

AI has entered education faster than most systems can govern it. For displaced learners, that speed creates a tempting story: if there are not enough schools, teachers or textbooks, perhaps a chatbot can fill the gap. It cannot.

This does not mean AI has no role in refugee education. It means the order matters. A useful tool must sit inside a real learning system. That system still needs a curriculum, trusted adults, safe places to learn, reliable ways to assess progress and a pathway to a recognized next step.

UNESCO put that question at the center of Digital Learning Week in September 2026. Education leaders examined AI tutors, autonomous systems and synthetic content while asking who retains agency, how young people develop and how education remains accountable to the public. The message was not to reject technology. It was to keep human agency, equity and the public purpose of education at the center.

The central pointAI may help a learner understand a lesson. It cannot, by itself, guarantee that the learner has a device, affordable data, a safe environment, a qualified educator, valid assessment or a route into further study.

Connectivity is the first test, not a footnote

The International Telecommunication Union reported in its 2025 global estimates that almost three-quarters of the world was online. That progress is real. So is the remaining divide: 2.2 billion people were still offline, mainly in low- and middle-income countries. ITU also warned that coverage alone does not settle the problem because quality and affordability gaps persist.

These are global figures, not refugee-specific estimates. They still expose a basic design failure. Any learning model that assumes uninterrupted broadband, a personal laptop and a quiet room will exclude people before the lesson begins. Refugee learners may share devices, depend on prepaid mobile data, move between neighborhoods or countries, work during the day, or lose access without warning.

Global internet use shows the access gap

Approximate share of the world population, 2025

ONLINE ~74%

OFFLINE ~26%

2.2Bpeople remained offline worldwide

Unequalquality and affordability persisted even where coverage existed

Source: ITU, Measuring Digital Development: Facts and Figures 2025. Percentages are rounded from ITU’s description that almost three-quarters of the global population was online.

Responsible digital education must therefore work on the learner’s actual device, tolerate weak connections, support downloadable or asynchronous work and provide a human route back into learning when technology fails.

AI should strengthen the teacher’s reach

On September 16, 2026, UNESCO published a discussion with 700 students and education experts about learning with new technologies. The students asked practical questions: how can they use AI for homework without losing the habit of learning, which sources can be trusted and what happens when teachers use AI?

The strongest answers were human ones. Students still need effort, curiosity, source verification and collaboration. Teachers should retain control over the tools they use. AI can offer guidance, but education cannot become a transaction between a child and an algorithm.

That distinction is especially important in displacement. A teacher or trained facilitator does more than explain content. That person notices when a learner is missing, confused, unsafe, isolated or carrying responsibilities that interfere with school. Technology can help organize information. It cannot fully read a family situation, build trust across cultures or decide when a learner needs protection rather than another exercise.

Five conditions for responsible AI-supported learning

Drawing from UNESCO’s human-centered approach and the realities of low-connectivity learning, KnowArk believes five conditions should guide any use of AI with displaced learners:

Teacher oversightAI supports educators and facilitators. It does not make high-stakes decisions about a learner on its own.

Verified curriculumGenerated explanations must connect to defined learning goals, credible sources and clear assessment standards.

Low-bandwidth accessLearning should remain usable on shared or basic devices and recover gracefully after an interrupted connection.

Safeguarding and privacyChildren should not have to surrender unnecessary personal data to learn. Human review and clear escalation routes are essential.

Local human supportTrusted partner centers, educators and community mentors give online learning a physical point of continuity.

Why hybrid learning matters after COVID-19

The pandemic changed how the world understands online education. It proved that learning can continue beyond a school building. It also proved that putting content online is not the same as creating access.

Hybrid learning is valuable because it does not force a false choice between digital scale and human support. Online materials can give learners flexibility when transport, documentation, work or family obligations interrupt attendance. Local partner centers can provide devices, connectivity, facilitation, safeguarding and a place where learners are known.

This is the model KnowArk is developing: a flexible K-12 pathway that begins with the urgent secondary-school gap, supported by online learning and trusted local centers. The approach is intended to keep routes toward high-school completion, English learning, college preparation and technical or vocational education open. Program availability, recognition arrangements and local delivery will depend on the specific pathway and partner context.

KnowArk’s position: AI should reduce friction for teachers and learners. It should never be used to disguise the absence of teachers, connectivity, safeguarding, valid assessment or a credible educational pathway.

The right question is not whether to use AI

The better question is what kind of educational relationship the technology serves.

If AI simply gives answers, it may weaken the work of learning. If it helps a teacher translate an explanation, identify a misconception, create additional practice or support a learner between sessions, it can be useful. If it requires constant high-speed access, collects unnecessary data or pushes a child through content without meaningful human contact, it can deepen the very inequality it claims to solve.

Refugee education does not need another shortcut that works mainly for people who already have stable access. It needs systems designed from the start for interrupted lives, uneven connectivity and learners whose education must remain portable without becoming impersonal.

Help build the human layer around digital learning

KnowArk welcomes refugee-serving organizations, educators and community leaders interested in inclusive hybrid learning pathways.

Explore a partner-center conversation Review KnowArk’s learning pathways

Sources and methodology

This article distinguishes published external evidence from KnowArk’s perspective and model in development. The connectivity visualization uses a rounded global estimate and is not a refugee-specific measurement. No learner outcomes, enrollment totals, completed partnerships or accreditation claims are presented.

UNESCO, The Algorithm in the Room, published September 8, 2026 and updated September 9, 2026.
UNESCO Digital Learning Week 2026, held September 8 to 11, 2026.
UNESCO, Learning with New Technologies and Artificial Intelligence, published September 16, 2026.
ITU, Measuring Digital Development: Facts and Figures 2025.