How to Cite: Romero Sánchez, Diego, Diana Herrera Duque and Omar Campos Rivera. "A Pathway Through the Study of Contemporary Latin American Architectural Theories: From Traditional Qualitative Methods to AI". Dearq no. 42 (2025): 31-38. DOI: https://doi.org/10.18389/dearq42.2025.04
Diego Romero Sánchez
Universidad Nacional de Colombia
Diana Herrera Duque
Universidad Nacional de Colombia
Omar Campos Rivera
omar.camposrivera@manchester.ac.uk
The University of Manchester, UK
Received: January 31, 2024 | Accepted: January 19, 2025
This text examines the methodological development used to explore a field that remains under-researched in the twenty-first century: Latin American architectural theories. Addressing this topic is essential, given the vast array of dissemination materials and the thematic, contextual, and ideological differences that characterize them. To address this, we outline the transition from a traditional qualitative research approach to the development of an artificial intelligence model. This process includes an explanation of the analytical framework—comprising seven approaches—the selection of academic journals as primary sources, and the validation of both through a natural language processing model.
Keywords: Latin America, digital humanities, research methodologies in architecture, artificial intelligence, contemporary architectural theory.
The interest in Latin American architectural theory, which gained momentum through rigorous scholarly work in the late 20th century, still grapples with the challenge of integrating architectural thought into the broader context of contemporary global issues. Despite the extensive body of research, much of it has largely focused on specific topics. In this context, we ask: What characterizes contemporary architectural theory in Latin America? Acknowledging the complexity of this question, we employed various methodological approaches that, while proving insufficient, also revealed the vast breadth of architectural knowledge available—often diverging from the traditional understanding of theory, which is primarily linked to design practice. This realization led us to explore alternatives capable of integrating a sufficiently broad dataset to validate or challenge our hypotheses. We opted for content analysis using Machine Learning (ML), which enables us to process and analyze large volumes of unstructured natural language data (NLP) to construct a comprehensive overview of contemporary Latin American architectural theory.
Thus, this text outlines the methodological trajectory by breaking down the overarching research question into multiple inquiries, focusing on the state of knowledge, the construction of an interpretative framework, and, the initial implementation of a methodological design that integrates Artificial Intelligence (AI) processes.
Architectural theory in Latin America has been, to a great extent, shaped by the Seminarios de Arquitectura Latinoamericana (SAL), from their inaugural edition1 in Buenos Aires in 1985 to the most recent gathering in Castro, Chiloé, in 2024. The SAL have served as a critical arena for the construction of Latin American architectural thought, fostering discussions on the relationship between identity, history, and modernity within a specific geographical and cultural context. Central to these debates has been the assertion of autonomous discourses and analytical frameworks, tailored to the region's unique architectural and socio-cultural conditions (Zambrano 2015, Ramírez 2013). The theoretical reflections emerging from the SAL found their most significant expression in the works of Marina Waisman, particularly El interior de la historia (1990) and La arquitectura descentrada (1995). These books remain among the last comprehensive texts to systematically engage with the architectural theory of their time. While this body of theoretical work explicitly seeks to critically reflect on Latin American architecture, it operates within the broader disciplinary framework of architectural theory—a field that, according to Waisman (1990), serves three fundamental purposes: constructing historiography, engaging in project criticism, and informing design. In other words, architectural thought is inherently project-driven, positioning the architectural project itself as a form of theoretical inquiry.
Since then, the discourses and challenges of Latin American architecture have expanded and become more complex, alongside its evolving professional horizons. This transformation has allowed architecture to venture into interdisciplinary fields and engage with global contextual issues, such as social and environmental crises, the rise of the internet, and emerging technologies. While extensive debates worldwide have questioned the state of contemporary architecture and its theoretical foundations, in Latin America, it is no longer possible to speak of a single, unified theory—rather, there are many. Despite the persistence of a more conservative, Eurocentric perspective that upholds the notion of a singular, discipline-bound architectural theory—primarily focused on design theory (Schumacher 2010, Armesto 2000)—this position has faced substantial critique. Counterarguments emphasize the impossibility of maintaining a singular theoretical model for a profession that has expanded both its scope of action and conceptualization. As a result, alternative architectural theories have emerged, integrating debates rooted in social contexts. These approaches demonstrate that theoretical frameworks beyond the realm of design are not only valid but also essential and fully operative (Torre 1996, Haddad & Rifkind 2014).
Within this scenario, new perspectives—such as decolonial approaches—have entered the architectural discourse, challenging the notion of a singular, universal theory. These perspectives argue that accepting a single theory ultimately means endorsing a European theory—an extension of the colonial, racist, and exclusionary world-system. Thus, they advocate for the acknowledgment of epistemic diversity (Lara 2020), particularly when reassessing architectural history. This growing complexity has also been recognized in numerous revisions of Latin American architectural projects (Hernández 2005, Segawa 2005, del Pino & Carrión 2021). However, none of these studies have explicitly examined the current state of architectural theory itself.
As a result, the current overview reveals a paradox: while the question of theory is rarely addressed directly, a vast array of places of enunciation, emerging dissemination dynamics driven by digital media, and intricate actor networks have reshaped where theory can be found. In this context, embracing the diversity of knowledge in a transversal manner becomes a pressing challenge—one that seeks to avoid reductionism and homogenization. From this perspective, the central hypothesis of this study is that architectural theories can be understood through different epistemic frameworks, which, in turn, can be categorized as approaches to contemporary Latin American architectural theory. However, researching this plurality requires new investigative models—both operational and ethical. If in the past, the process of constructing theory was not a primary concern, today, amid an explosion of perspectives and theoretical outputs, this very process has become fundamental to theory's legitimacy.
Given the challenges posed by the contemporary scenario of architectural theory, developing a methodology that recognizes these conditions becomes a research outcome in itself. The process of constructing a framework for analyzing information and mapping the theoretical panorama has evolved from exploratory approaches rooted in researchers' experience to more structured systematizations, ultimately leading to the implementation of artificial intelligence for specific analytical purposes. However, reaching the stage of AI-assisted analysis marks the final step in a long process, and its application does not imply a positivist stance on theoretical research.
The study initially began with broad regional overviews. This preliminary exploration revealed not only vast amounts of information to be analyzed but also its inherent diversity and the potential for categorizing architectural theory thematically or ideologically. However, at this stage, a major limitation emerged: the difficulty of interpreting these theoretical positions relationally. As a result, establishing a systematized analytical framework was still premature, as no solid interpretative model had yet been developed.
These panoramic investigations were conducted using a combination of qualitative research methods, reflecting the study's exploratory nature. This phase included in-depth interviews, mapping of think tanks and intellectual centers, identification of actor networks, archival research on awards and biennial documents, and content analysis of frequently cited texts, among other strategies. The initial dataset provided insights into potential interrelations among key actors and helped clarify the thematic and ideological approaches embedded in Latin American architectural discourse.
The previous stage confirmed the assertion that there are indeed multiple architectural theories, where engaging with theory means engaging with spatial thinking related to the construction of the built environment. Given this diversity of thought, employing various categories of analysis becomes essential. Thus, the challenge shifts from merely identifying trends to categorizing them, in order to understand the thematic and political stances shaping knowledge production. While the categories proposed by Crysler, Hilde, and Cairns (2011), or by Zaera-Polo (2016), align with certain Latin American patterns, their direct application to the region reveals limitations—some topics appear redundant, while others remain insufficiently explored. This highlights the strong relationship between thematic concerns and ideological positions.
In this context, the second phase of research focused on thematic relationships as a systematic approach to understanding the information and identifying the specific theories in question. In other words, a theoretical framework was proposed to analyze the evolution of certain thematic trends in Latin America. This framework outlines seven approaches that are not mutually exclusive but rather represent both ethical positions and theoretical devices used to analyze architecture and the processes of constructing habitable space in the region. Some of these approaches are situated—specific to Latin America—while others address global issues. Their construction involves a continuous exchange between the information sources and analysis, guided by common points of intersection. Despite the challenge of conceptualization, this process establishes an initial analytical dimension that underscores the region's epistemological diversity.
Building upon these theoretical frameworks, the next stage focused on data collection and analysis. At this point, the debate centered primarily on selecting the units of analysis—defined as the objects of study that would validate the theoretical framework's hypotheses. The process began with a pilot study using discourse analysis, approached from a descriptive statistical perspective. This pilot was designed using a sample of articles from three Colombian academic journals: Bitácora, Revista de Arquitectura, and Dearq. These journals were selected based on prior familiarity, peer-review standards, and the availability of a sufficiently extensive archive within the defined timeframe (2000–2023). The goal was to assess whether the theoretical approaches could effectively classify all possible contents within the discourse. Simultaneously, a network analysis was conducted to identify knowledge hubs, focusing on think tanks and research centers that bring together different actors in architectural discourse. These centers were classified according to theoretical frameworks and georeferenced to verify the findings of the initial stage. Both exercises proved valuable in testing the operability of the proposed theoretical frameworks. However, they also revealed a major limitation: the sheer volume of content that could be categorized within the scope of contemporary Latin American architectural theory far exceeded the capacity for manual review.
Thus, while these approaches serve as valuable analytical tools, the real challenge lies in constructing a comprehensive panorama—one that integrates the broadest possible dataset to fully capture the geographic and conceptual diversity of Latin American architectural thought.
Given these limitations, addressing this question requires an explanation of how this research engages with digital humanities. In architecture, such approaches have been explored in works such as the Black Architects Archive (Cephas 2020) and The Displacement of Architecture in Postdigital Humanities: Neoanalogue Indexes, Syntaxes, and Configurations (Ioannidis 2018), where large collections of architectural works are analyzed and categorized for digital visualization. A case more directly related to the context of this research is the study of the archives of the Quito Architecture Biennial (Zambrano et al. 2021), which examines the content to reveal the underlying logics of knowledge dissemination. However, these perspectives do not encompass large-scale categorizations of complete texts across Latin America, nor do they seek to extract theoretical content.
In this context, the research proposes the use of two tools: the development of a robust database that consolidates diverse outputs, the individuals involved in these epistemological processes, and the architectural projects themselves; and an NLP-based analysis of the elements identified within this database, using the artificial intelligence platform Dataiku. This type of analysis functions by feeding the platform with key concepts, allowing it to detect their presence even when they do not appear in the exact words with which they were originally defined.
While these tools are complementary, they enable large-scale analyses that were tested in the pilot studies. On one hand, the database provides a foundation for a descriptive quantitative analysis, such as the geographic concentration of architectural production segmented by thematic categories, social network analysis to identify key figures in the dissemination of ideas and thematic clusters, and chronological tracking of theoretical approaches. On the other hand, artificial intelligence poses a challenge by attempting to conduct content analysis based on text encoding, enabling the identification of complex ideas and their relationships with different theoretical perspectives. This means that AI facilitates determining which theoretical approaches are present in a given text, but applied at a large scale across architectural discourse. Processing this data not only enables the classification and grouping of information according to previously proposed theoretical frameworks but also makes it possible to establish new categories or clusters based on emerging trends or identified characteristics. It also contributes to an understanding of how spatial and geographic ideas are articulated from a Latin American identity perspective.
After constructing the database, key countries in theoretical production—Brazil, Argentina, Chile, Mexico, and Colombia— were identified based on the number of research centers, scholars, academic publications, and dissemination platforms. Following this, a selection of texts was defined for NLP-based analysis. The corpus for this exercise consists of academic journals, chosen for several reasons. Primarily, they represent the first stage of engagement with contemporary debates before reaching broader dissemination. Additionally, they encapsulate processes emerging from research centers and intertextual relationships. The selection criteria for the journals included: peer review, to ensure a minimum standard of quality; continuous publication for at least the last fifteen years (ideally since 2000); a broad theoretical scope, avoiding specialization in a single perspective (for instance, journals focused exclusively on urban processes would align more with developmentalism, environmental issues, and alterities); open-access availability; and publication by universities previously identified as key thought centers for contemporary theoretical discourse.
The final selection included Bitácora from Mexico, Dearq from Colombia, Arquitextos (Vitruvius) from Brazil, and ARQ+ from Chile, comprising a total of 435 issues and approximately 6,000 articles.
Since some older texts were only available in physical format, they had to be digitized using Optical Character Recognition (OCR)2 techniques to convert images into plain text for analysis. To process the content of the texts, a classification system was implemented using keyword identification (monograms or bigrams3) associated with each theoretical approach, a method commonly referred to in data analysis as 'ontology4.' These keywords include terms explicitly linked to theoretical frameworks, specific concepts, verbs, and the names of relevant authors—selected based on their significance for each approach. The ontology, built upon the conceptual definitions of the theoretical frameworks, serves to identify the prevalence of one or multiple perspectives within a given text. However, defining this ontology presents a key challenge pertaining to this research, as certain terms appear to be exclusively associated with a single theoretical framework yet also carry everyday meanings. For instance, the term form, when used in the context of disciplinary autonomy, refers to a specific architectural concept, yet in general language, it can also mean way (this ambiguity is particularly relevant in Spanish, though not in English). At a large scale, the ontology enables the visualization of how the presence of these theoretical frameworks fluctuates over time and how thematic and conceptual trends are represented within the corpus. Thus, the reliability of this method ultimately depends on the rigor of the ontology.
The initial analysis of the selected journals allows for key methodological reflections, including the validity of theoretical categories, the ways in which theory responds to social contexts, and the evolving relationships between different theoretical approaches over time.
The results confirm that the theoretical frameworks correspond to the defined timeframe, as the semantic values assigned in the ontology appear consistently across all journals. None of the categories tend to zero, nor do they reach negligible frequency levels— both of which would indicate misalignment with the themes discussed in the articles. Figures 1 and 2 (Dearq and Bitácora, respectively) illustrate how all categories are consistently recognized across different years. While some categories—such as disciplinary autonomy—exhibit higher statistical incidence, their values are not constant. This confirms that no single framework dominates the entire dataset or absorbs the terminology of others. However, achieving this balance required significant refinement, as initial ontologies exhibited strong bias toward certain dominant terms. Following these corrections, it is now possible to confirm that these frameworks not only exist but also effectively describe specific moments in architectural theory.
Regarding the validation of journals as sources of contemporary discourse, it is evident that their content aligns with the socio-political dynamics of their respective countries. Variations in certain theoretical frameworks correlate with specific historical events, demonstrating that thematic categories reflect shifts in architectural discourse at key moments. For instance, in Colombia, 2016 was marked by an intense national debate on peace, following the signing of the Peace Agreement between the FARC-EP and the Colombian state. This is reflected in Dearq, where themes such as environment, developmentalism, preexistences, and alterities—approaches related to political, social, and environmental issues central to the Peace Agreement—gained prominence, particularly in issues 18 and 19. A similar trend is observed in Chile with ARQ, where the 2019 social uprising triggered a spike in categories related to social processes and urban significance, such as preexistences and alterities. These findings validate the selected journals as responsive sources that reflect temporal evolution while also confirming that some categories are more context-dependent than others—meaning that their chronological fluctuations are more relevant than long-term trends.
Finally, the statistical behavior of each theoretical framework reveals conceptual affinities and contrasts. Some frameworks exhibit direct relationships, while others show inverse correlations. Alteridades and desarrollismo tend to move in parallel, albeit with different magnitudes. On the other hand, disciplinary autonomy, which posits a certain timelessness in the value of architecture, tends to remain dominant across all journals. However, there are moments when its prevalence decreases, giving way to frameworks with stronger social characteristics, such as desarrollismo in Dearq (Figure 1) or environment and alteridades in Bitácora (Figure 2). These findings highlight the need for the next stage of research to focus on evaluating the relationships between theoretical frameworks, as interactions between them are to be expected.
Figure 1_ All approaches in Dearq (Colombia) between 2007 and 2023. Source: the authors.
Figure 2_ All approaches in Bitácora (Mexico) between 2010 and 2023. Source: the authors.
To think about methodology is to think about theory. In a context where traditional methods of theoretical production are no longer viable due to the overwhelming volume of information, the use of artificial intelligence tools for critical analysis is not just desirable—it is essential. However, this implementation must stem from a critical thought process by researchers, where AI serves as a tool for epistemological development rather than an autonomous generator of knowledge. Its outcomes are shaped by how researchers employ it. For instance, ontologies are built based on the criteria and biases of those who design them, while the results they generate serve as new inputs for analysis.
Constructing a custom theoretical framework helps clarify complex issues, such as the diversity of contemporary thought and the role of interdisciplinarity in architectural discourse. Analyzing architectural theories through theoretical frameworks allows for a deeper understanding of the epistemological diversity of Latin American architecture, particularly in relation to the expanding scope of the profession—one that no longer revolves solely around architectural design in the traditional sense. Today, the architectural project operates within a broader realm of thought and the construction of habitable space.
Additionally, the proposed theoretical frameworks reveal varying implications for interdisciplinarity. Some underscore Latin America's closer alignment with socially driven issues—whether political, historical, or sociological—while others emphasize aesthetic and technical concerns within the discipline. The interactions between these approaches remain a subject of ongoing research, and further exploration will deepen our understanding of their distinctly Latin American character.
Beyond confirming existing perceptions, AI-based models also validate diverse sources. Once refined, these models can be expanded to analyze other contemporary theoretical sources—such as podcasts, conferences, social media, and exhibitions—leveraging metadata to construct a more comprehensive picture of contemporary architectural discourse.
The research journey outlined here demonstrates that artificial intelligence is not only inevitable but also highly productive in contemporary academic research. In this study, AI has proven instrumental in identifying processes of territorial identity construction, integrating Latin America's social and cultural characteristics, and offering new insights for schools of architecture across the region.
1 Even though previous works about Latin American architecture developed in other geographical contexts could be considered as precedents, Ramírez (2024) points out that Marina Waisman was the first to build specific knowledge focused on understanding Latin American architectural production.
2 Often journal's files from early 21st century are available only in image formats or PDF which are not immediately recognized as plain text (as the one use in apps like Notepad).
3 Monogram is a keyword composed by only one word (for example, 'inhabitate'). Bigram is composed by two words, such as 'built heritage'.
4 In this case, the term is not used in the philosophical and epistemological sense. The word is used to describe the classification system employed by AI.