AI in Education: Lessons from Brazil and the Global South

Written By Ana-Paula Correia
AI in Education: Lessons from Brazil and the Global South

Amazon River near Manaus, Brazil (Photo by Nareeta Martin on Unsplash)

This month, I had the privilege of visiting Brazil, where I participated in a conference in Curitiba and presented at the University of São Paulo. Both engagements allowed me to share my research on Artificial Intelligence (AI) in education and teacher education. Yet the greatest lessons came not from what I presented, but from what I heard and witnessed.

In one particularly moving session, Professor Josemir Almeida Barros from the Universidade Federal da Rondônia shared his research on rural education in the Amazônia ribeirinha, the riverside communities of the Amazon. His presentation highlighted the everyday realities of teachers working in precarious conditions: schools with minimal infrastructure, frequent power outages, scarce educational resources, and unreliable internet access.

The stories were vivid. Students and teachers travel long distances, sometimes hours by boat, through rivers where they must navigate floods, storms, intense heat, and even threats from alligators and jaguars simply to reach school. And yet, learning happens. These educators embody resilience and compassion, teaching multilevel classrooms while also learning from students and families about the rhythms of the river, survival strategies, and ways of adapting to a harsh but life-sustaining environment.

Listening to Professor Josemir Almeida Barros’ testimony, I was struck by the humility and courage of these teachers. Then, when it was my turn to speak about AI in education in the context of the United States, the contrast was undeniable. How can we speak of generative models and personalized learning dashboards in contexts where even electricity is a luxury?

This tension raises an urgent question: Artificial intelligence is reshaping education worldwide, but is it doing so in an equal way?

Artificial Intelligence at a Crossroads: Global Aspirations, Different Realities

UNESCO’s 2025 eBook, AI and the Future of Education: Disruptions, Dilemmas, and Directions, captures this global dilemma. AI holds transformative potential to expand access, reduce administrative burdens, and provide adaptive support. Yet one-third of humanity remains offline, and access to cutting-edge systems depends on infrastructure, subscriptions, and linguistic privilege.

Recent research confirms these concerns (Ifenthaler et al., 2024). These authors’ Delphi study with global experts identified three priorities for AI in education: privacy and ethical use of data, trustworthy algorithms, and equity. These are not abstract concerns; they are lived realities for millions of learners and teachers. Without intentional action, AI risks reinforcing the very inequalities it promises to solve.

Confronted with the realities in Brazil, I saw this was not a theoretical policy challenge; it was a human challenge. AI cannot be discussed apart from the lives of teachers who teach with intermittent electricity, or learners whose only path to school is a long journey by river. If AI is to serve education globally, it must be built on an ethics of care, relational teaching, strong governance, and a commitment to inclusion.

Adopting a “Care and Compassion by Design” Framework

Ethics cannot be an afterthought. AI in education must be designed from the ground up with compassion and empathy. UNESCO (2025) calls for policies that prioritize human rights, justice, and inclusion over narrowly focused commercial interests. Ifenthaler and colleagues (2024) argue for a “value-centered design” approach, which requires the identification and aggregation of all stakeholders’ values and ensuring system functionalities are linked to ethical principles.

A framework of “care and compassion by design” translates into seven shifts (UNESCO, 2025):

  1. Embedding fairness, accessibility, and inclusivity as design principles.

  2. Anticipating vulnerabilities and preventing harm.

  3. Ensuring transparency so that educators and learners understand how AI systems make decisions.

  4. Supporting teacher agency rather than replacing it.

  5. Creating spaces for local adaptation and cultural relevance.

  6. Designing for low-bandwidth and offline use.

  7. Ensuring that AI systems enhance, not erode, human dignity.

Care-centered design is strategic and practical. It acknowledges that education is relational and political, and that technologies built without compassion risk exclusion, mistrust, and failure.

Revaluing Human Teachers and Relational Pedagogy

A central theme across both the UNESCO eBook (2025) and the Delphi study (Ifenthaler et al., 2024) is that AI must augment, not replace, human teaching. Education is not reducible to prediction models or automated tutoring systems. It thrives in relationships.

Pedagogy First. Every AI adoption must begin with a pedagogical question: What do we want learners to become? Tools should be chosen only if they support the answers to this question.

Reaffirming Teachers. AI systems should serve as tools that extend teachers’ abilities, not undermine their expertise. Teachers remain designers of learning, capable of cultivating autonomy and dialogue.

Preserving I-Thou Encounters. Drawing on Martin Buber’s (1932) relational pedagogy concept, education must preserve the I-Thou relationship, genuine encounters of recognition and growth. It prioritizes building human connections before the exchange of information, recognizing both teacher and learner as unique beings engaged in mutual growth. Machines cannot replicate this.

Resisting Isolation. While personalization can support learners, hyper-personalization risks isolating them in echo chambers. Instead, AI should foster collective intelligence, peer-to-peer collaboration enriched by human-to-machine dialogue.

As Ifenthaler et al. (2024) highlight, protecting pedagogy is a global concern. “Pedagogy with AI” was rated among the most pressing challenges, with experts warning that technological enthusiasm must not overshadow human learning and development goals.

Implementing Ethical Governance and Policy Guardrails

Policy is where ideals become actionable. Both UNESCO (2025) and Ifenthaler et al. (2024) agree that effective AI integration depends on the establishment of clear governance frameworks. This begins with AI literacy, which should be considered as foundational as reading and mathematics. Such literacy must encompass not only conceptual understanding of how AI works, but also critical awareness of where bias may appear and creative capacity to use AI ethically and responsibly.

Equally crucial is the development of trustworthy algorithms. Ifenthaler et al. (2024) stress that transparency and fairness are essential to building confidence in AI systems. Without trust, adoption will inevitably falter. Policies must also exercise ethical foresight, moving beyond narrow risk management to create enabling conditions for innovation and collaboration. At the same time, governance should be adaptive, designed to evolve iteratively through developmental experimentation and collective sense-making. Finally, the integration of AI must remain evidence-informed, carefully selective, and justified by tangible benefits—whether improving learning outcomes, reducing teacher workload, or promoting fairness.

These steps cannot be accomplished in isolation. They require coordination among policymakers, technologists, educators, and communities. Just as importantly, they demand humility. Policies must be shaped through dialogue with those most directly affected, particularly in the Global South, where the consequences of ill-conceived implementation are most deeply felt.

Confronting Coded Inequalities and Ensuring Inclusion

Perhaps the greatest challenge is that AI systems inherit and amplify societal inequities. As Baker and Hawn (2021) show, algorithmic bias can distort predictions and reinforce exclusion. Confronting these “coded inequalities” requires deliberate action.

Center Local Voices. Implementation must begin with context. Tools co-created with educators, learners, and communities are more likely to reflect different values and knowledge systems. Participatory design cycles are essential.

Address Linguistic Bias. Over 99% of the world’s languages are not represented in large AI models. This situation risks leaving out other forms of knowledge (e.g., indigenous knowledge). Solutions may include designing multilingual models and advocating for open-source alternatives.

Invest in Infrastructure. AI cannot substitute for systemic investment in infrastructure, devices, and connectivity. Without these, adoption will fail, or worse, deepen inequity.

Layer Human Support. AI must be layered with culturally relevant pedagogy and human support. As UNESCO (2025) notes, technology cannot replace the empathy and contextual knowledge that only teachers bring.

Ifenthaler et al. (2024) make it clear that experts ranked equity and fairness among the top three challenges in AI for education. Without intentional inclusion, AI will replicate rather than remedy systemic exclusion.

Supporting Teachers in the Global South

While in Brazil, I was reminded that teachers are not simply implementers of policy; they are the very foundation upon which education rests. Supporting them must be our priority.

This means expanding teacher education programs that cultivate AI literacy and fluency while staying attentive to local realities. It also means ensuring that infrastructure investments ease teachers’ daily burdens instead of adding new layers of complexity. Just as important is the creation of cross-regional professional networks where educators can share practices, insights, and challenges, fostering a global community of knowledge. And when designed thoughtfully, AI systems can relieve teachers from time-consuming administrative tasks, giving them the freedom to devote more energy to what they love: teaching.

Yet, the flow of support cannot be one-directional. Educators in the Global North must adopt a more humble approach, acknowledging that resilience, creativity, and resourcefulness are qualities they can and should learn from their colleagues in the Global South.

Ultimately, supporting teachers is not only about equipping them with AI tools but also about honoring their wisdom and experiences. When we value teachers as partners in shaping the future of education, we not only strengthen education systems but also reaffirm the profound human connections at the center of learning.

Beyond Innovation: Designing for Necessity

As we envision the future of AI in education, with generative models that adapt in real-time and dashboards that personalize learning at scale, we must also reckon with the stark reality that for many communities, even electricity is a luxury. As the old saying goes, necessity is the mother of invention. Low-bandwidth and offline solutions, such as SMS-based tutoring, lightweight apps that sync when possible, and dashboards optimized for intermittent connectivity, demonstrate ingenuity when resources are limited and more of that is needed.

By embracing these constraints, we broaden, not narrow, the horizon of educational technology. The challenge is not simply to innovate for the best-connected classrooms but to ensure that AI serves learners everywhere, especially those for whom access is most difficult. Only then can we speak credibly of equity in the age of intelligent systems.

References

Please cite the content of this blog:

Correia, A.-P. (2025, September 30). AI in Education: Lessons from Brazil and the Global South. Ana-Paula Correia’s Blog. https://www.ana-paulacorreia.com/blog/ai-in-education-lessons-from-brazil-and-the-global-south

Ana-Paula Correia
Previous

The Maestro’s Baton: On Leadership, Learning, and Vision
Next

Artificial Intelligence, Silence, and Spiritual Growth: Reflections from Fátima