Win-win situation: Generative AI in Educational Constrained Environments

• Paulo C. R. Pinho UFPel
• Renan Z. da Silva UFPel
• Tiago T. Primo UFPel
• Diego Dermeval UFAL
• Ig I. Bittencourt UFAL
• Seiji Isotani USP

DOI: https://doi.org/10.5753/sbie.2025.12497

Resumo

The global digital divide remains a major barrier to equitable access to technology, and this scenario compromises the pedagogical use of technology and limits learning opportunities for students from disadvantaged communities. Inequality in access to technology is a global issue that directly affects educational and socioeconomic development, especially in countries like Brazil, where there is a significant disparity in access to the Internet and technological devices between different regions and social classes. Despite ongoing efforts, 6% of the public schools still lack broadband, and 3.3% remain without electricity, according to the Ministry of Communications. This gap is most severe in rural and undeserved communities, where limited infrastructure and resources further hinder quality education. Traditional educational technology models are inadequate in such contexts. To address this challenge, we propose Unplugged Artificial Intelligence in Education (AIED), a framework designed to deliver accessible educational technologies without relying on modern infrastructure, stable internet or advanced digital skills. Our goal is to bridge the digital divide and empower students and educators with innovative tools that promote inclusion and reduce educational inequalities. This study presents a model for the deployment of LLMs in offline environments to support literacy and learning in resource-constrained areas. The proposed solution is low-cost and adaptable, allowing schools without internet access to benefit from generative AI. By deploying LLMs locally, the system provides real-time language support and generates educational content without requiring an internet connection. Preliminary results demonstrate the feasibility of running optimized generative models locally, offering practical insights into AI-powered solutions for low-resource environments. By integrating AI agents with curated educational resources, this approach promotes equity and sustainability, ensuring that generative AI serves as a tool for inclusion and innovation in education.

Referências

ANATEL (2024). Painel conectividade nas escolas. Disponível em: [link]. Acesso em: 20 jun. 2024.

CNE (2020). Parecer cne/cp nº 11/2020, orientações educacionais para a realização de aulas e atividades pedagógicas presenciais e não presenciais no contexto da pandemia. Disponível em: [link] parecer-cp-2020. Acesso em: 04 ago. 2025.

Isotani, S., Bittencourt, I. I., Challco, G. C., Dermeval, D., and Mello, R. F. (2023). Aied unplugged: Leapfrogging the digital divide to reach the underserved. In International Conference on Artificial Intelligence in Education, pages 772–779. Springer.

Radford, A., Narasimhan, K., Salimans, T., Sutskever, I., et al. (2018). Improving language understanding by generative pre-training. OpenAI.

Rodríguez, M. R. B. and Cobo, C. (2022). Covid-19 and education in the global south: Emergency remote learning solutions with long-term implications.

UNESCO (2023). Global education monitoring report summary, 2023: technology in education: a tool on whose terms? Disponível em: [link]. Acesso em: 25 ago. 2025.

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., and Polosukhin, I. (2017). Attention is all you need. Advances in neural information processing systems, 30.

Wang, B., Liu, J., Karimnazarov, J., and Thompson, N. (2024). Task supportive and personalized human-large language model interaction: A user study. In Proceedings of the 2024 Conference on Human Information Interaction and Retrieval, pages 370–375.

Yin, W., Xu, M., Li, Y., and Liu, X. (2024). Llm as a system service on mobile devices. arXiv preprint arXiv:2403.11805.