Search
Browse By Day
Browse By Time
Browse By Person
Browse By Room
Browse By Committee or SIG
Browse By Session Type
Browse By Keywords
Browse By Geographic Descriptor
Search Tips
Personal Schedule
Change Preferences / Time Zone
Sign In
Education must ensure success for all students, regardless of their backgrounds, experiences, or identities. As society becomes more diverse and interconnected, these changes are reflected in classrooms, making it essential for education systems to prioritize inclusion. By embracing the various aspects of diversity—cultural, economic, gender, special needs, and beyond—education can guarantee that every student is supported, valued, and positioned to succeed both academically and personally. The curriculum, a core component of education, should embody these principles of diversity and inclusion, with textbooks clearly reflecting them to guide respectful and representative learning for all students (Organización de las Naciones Unidas para la Educación, 2017; Cerna et al., 2021; Fuchs et al., 2018).
To achieve this, it is critical to recognize the broader role that formal education plays in shaping societal norms and transmitting what is considered "legitimate knowledge" (Schissler, 1990). Education influences national identities, often through the selective emphasis on particular ethnic groups in a nation’s history. Schools, therefore, are not just educational institutions but also spaces where political understanding is formed, and where the politics of recognition are contested (Page et al., 2010; Apple, 2012). The direction of schooling is shaped by political decisions that influence curricula, learning materials, evaluation criteria, and budget allocations, ultimately mediating the transfer of political culture across generations.
Research has extensively explored how political actors may use educational institutions to pass their ideologies onto younger generations by shaping what content is deemed legitimate (Blumberg, 2008; Costin et al., 2020; Wallace et al., 2016; Pervez, 2020; Lee et al., 2023; Bromley et al., 2015; Normand, 2021). Textbooks, as a clear expression of such content, often reflect the sanctioned knowledge of their time, embodying perspectives endorsed by national governments and other influential groups. Disputes over textbook content underscore their importance as authoritative sources (Lerch et al., 2017). However, research in this area has been limited by small, inconsistent samples, often constrained by textbook availability and the challenges of using human coders and ensuring inter-rater reliability. Moreover, much of this research is concentrated in English-speaking countries, with limited focus on low- and middle-income countries (LMICs) (Vojíř et al., 2019). This is particularly concerning, as policies aimed at reducing inequality are often more difficult to implement in these economies (Programme, 2024).
This paper addresses three critical gaps in the literature. First, by focusing on Brazil, we bring the discussion of diversity and inclusion in textbooks and curricula, using computational methods, to low- and middle-income countries. Second, leveraging Brazil’s Programa Nacional do Livro Didático (PNLD), a nationwide textbook program covering all public schools, we overcome the limitation of small samples and potential biases by examining all textbooks available to public schools. We also extend our analysis by characterizing each textbook individually. While most studies analyze the entire corpus, we examine the variation in diversity and inclusion across individual books, allowing us to identify differences in how these themes are represented.
We analyzed all 102 humanities textbooks available to public secondary schools across Brazil. These textbooks are accessible to all 180,000 public schools in the country, reaching over 6 million students as part of the PNLD under the Brazilian Ministry of Education. The textbooks are evaluated by teachers nationwide in accordance with the National Common Curricular Base (BNCC). PNLD categorizes these books into two main areas: "Languages and their Technologies" and "Applied Humanities and Social Sciences." Languages and their Technologies include subjects such as Portuguese, physical education, music, theater, dance, and visual arts. Applied Humanities and Social Sciences cover subjects such as history, geography, sociology, and philosophy. As part of the PNLD, textbooks are provided to schools free of charge by the Federal Government. By characterizing all textbooks, we can match them to individual schools, offering a comprehensive view of education in the country, particularly regarding regional differences in access to more or less diverse and inclusive textbooks.
Critical Discourse Analysis (CDA) guides our examination of how language in educational texts can perpetuate biases, prejudices, and discriminatory practices. Through this approach, we identify marginalization, underrepresentation, and the reinforcement of dominant ideologies within these materials. CDA also emphasizes the importance of self-reflection in the creation and selection of textbook content, encouraging educators to challenge oppressive narratives and promote more inclusive representations (Mullet, 2018; Moore, 2022; Tabatadze et al., 2023).
Guided by this framework, we employ natural language processing (NLP) to analyze textbook content with a focus on diversity and inclusion. NLP enables us to identify patterns and trends that may be overlooked in manual assessments or that are impractical to process through reading or manual coding (Grimmer et al., 2011). This approach provides a clearer understanding of inclusivity in educational resources, contributing to a more equitable learning environment (Ferreira-Mello et al., 2019).
In this preliminary analysis, we replicated the core study by Lucy et al. (2020), examining gender and ethnicity depictions in Brazilian textbooks. Adapting the proposed NLP methods presented significant challenges, as key components were designed for English, requiring substantial modifications. These adaptations, which are an additional contribution to academic research, address the bottlenecks researchers face when applying NLP tools to non-English languages. By utilizing Portuguese-specific models, such as neuralmind/bert-large-portuguese-cased and spaCy 3.7, we successfully tailored the tools for analyzing Portuguese texts.
Furthermore, our access to all humanities textbooks used in Brazilian public secondary schools allowed us to analyze variations in diversity and inclusion across different textbooks. To the best of our knowledge, this approach—focusing on variations across textbooks rather than treating the corpus as a whole—is unique in the field.
Our results show disparities in representation that reinforce stereotypes and unequal narratives, with male and white individuals being more represented in Brazilian textbooks. The regional variation in book usage allows us to show correlations between textbook choices and regional characteristics. These results underscore the need for a thorough review of educational content to ensure it reflects the diversity of Brazilian society and supports an inclusive learning environment.