Large Language Models (LLMs) have revolutionized how we generate, interact with, and process language. Still, these models are biased toward WEIRD (Western, Educated, Industrialized, Rich, and Democratic) values. This bias is not merely linguistic but also cultural. Sociocultural contexts influence how people express ideas, interpret meaning, and communicate. In low-resource language settings, where data and cultural representation are limited, this issue becomes even more pronounced when models are applied without cultural adaptation, often leading to outputs that are irrelevant, inaccessible, or even harmful. In this paper, we argue for the importance of incorporating sociocultural context into LLMs. We review existing frameworks that explore culture in Natural Language Processing (NLP), and examine some work aimed at culturally aligning language models. As an illustrative scenario, we analyze the case of Guinea-Bissau. In this linguistically and culturally diverse country, Portuguese is the official language but not the primary means of communication for most of the population, highlighting the urgent need to adapt educational materials to the local sociocultural context. Finally, we propose a revised framework to address the challenge of adapting educational materials to diverse contexts, aiming to improve both the relevance and pedagogical impact of text adaptation.