MariannaCharitonidou’s ¹ book Architectural Drawings as Investigating Devices (2023) presents a survey and key insights into the recent history of representation in architecture. Marianna’s focus is through architects that challenged the discipline through key revolutionary drawings. With examples ranging from Le Corbusier’s Dom-Ino diagram to Mies’ collages, the book investigates insights into the Modern Movement. But it also includes certain tensions that make it relevant and interesting, for instance, John Hejduk’s axonometric projections experiments and Peter Eisenman’s formal axonometric diagrams are placed in contrast to Tschumi’s axonometrics. While Hejduk and Eisenman are presented as representational and formal, Tschumi’s drawings are rather political acts based on the semantics of the architecture program. 

The book inquires about the dialectical relationships between subject and object to disclose how architects problematized both architecture designs as well as the systems of representation through which they came about their projects. It investigates drawings as a form of research on representation, a unique form of cognition in architecture. The range of systems of representation studied in the book provide a means to access the understanding of the work of the architects investigated. Unique forms of representations give origination to unique architecture explorations; from thinking the drawings as means of research and exploration of the works of architecture, to thinking representation in terms of notation as buildings are built influenced by thinking through these systems. 

The book places in tension historical documents, sketches, diagrams, artifacts of representation, and drawings that transformed, revolutionized, and expanded canons defining what the architecture is, could be, or could have been. The book’s documentation includes unique archives that may influence architecture today. We have been researching representation alongside digital systems of representation correlating architecture and advances in computation. The immaterial mathematical vectorial quality of digital drawings has displaced architecture material embodiment through abstract relational diagrams. 

Back in the 2000s, Anthony Vidler ² discussed a genealogy of the architecture diagram in relation to computer representation through Paul Frankl,³ who activated a structural history independent from style by diagraming relationships in buildings. The diagram is often seen as a tool or a device to study relationships. While an architect could come up with a diagram as a synthesis, such as the LC Dom-Ino House (1914), the semantic functionality of diagrams to develop architecture could also be critiqued. Mies’ Barcelona Pavilion would not signify much represented as a diagram, abstract from its material medium embodiment such as its marble walls or its steel columns cladded with stainless steel. 

Diagrams can become too abstract as they become disembodied from both media and mediums, tending to represent linguistic thinking concepts and relationships translated into vague visual representations. Such translation may disregard the media used to represent content within a system of representation. The diagram is often in a continuous state of becoming, with neither media nor a medium through which it is expressing its apparent content. We question whether there is real content without a form of media or medium that defines the quality of a visual sign. This is the struggle of working through media. Media is composed by layers of systems of representation, which precede our drawings; analog to how, in linguistics, language is found to determine our thoughts. 

Jacques Derrida ⁴ critiques Ferdinand De Saussure’s equation ⁵ [sign = signified/Signifier or as we place it giving hierarchy to the Signifier over content: sign = Signifier (word-form-medium)/signified (concept-meaning-content)], establishing that linguistic signifiers induce actual content, presenting limitations to how we understand our reality. One could argue that in visual semiotics, the vanishing point in perspective or an axonometric angle as signs may become signifiers, inducing symbolic deterministic content or a media determinism in which “the medium is the message.” ⁶ Media both enable and separate subject-viewer experience and object-building design; and as a result spaceframes become active as a product of media. While as humans we think ideas, concepts, and our actions through language, we think architecture through visual representation, building up a social construct between subjects and spaceframes which, as a result, become loaded with ideologies of representation. 

We argue that if diagrams do not acknowledge media or a medium, they become semantically driven. And if this happens, their content may be extrinsically linguistic and not fully visual. The following examples develop a media- and medium-specific content, which in exchange revolutionized representation’s deep structures as defined by Erwin Panofsky.⁷ In Giuseppe Terragni’s Casa del Fascio (1932) there is an investigative complex emergent grammar as a visual syntax defines novel architectural signs out of three-dimensional spatial relationships. A diagram of Terragni’s building by Eisenman (1963–2003) can explain such complexity, although a reductive diagram cannot build up such complex grammar as the building does unless it engages with its material embodied medium of construction. 

Piet Mondrian’s Four Lines and Gray (1926) painting the 45° angle rotation of its canvas-frame disregarding any pictorial semantic external content, broke with the horizon in painting as a conventional semiotic sign and as a symbol. Hejduk’s Diamond House series or, in Project A, Axonometric 06 (1969) can be understood as a translation from painting into architecture of Mondrian’s experiments. The diamond houses present a rotation of the interior content-space in relation to the frame-envelope of the house. The envelope-façade of the house is in tension with its generative media, since the axonometric projection technique used can be read both as a 0–90° axonometric projection of the envelope but simultaneously also as a 45–45° axonometric projection of its interior content-space, making ambiguous the technique of projection in relation to the planimetric formal organization.⁸ 

In Stan Allen’s Axonometric Degree Zero (2019),⁹ Hejduk places at the same level of signification the architecture of a building design in relation to its deep structure generative media. For us, Hedjuk’s diamond houses axonometric projections displace their own system of representation, creating emergent ambiguous architectural signs such as the round columns that belong to both axonometric projective systems

(0–90° and 45–45°). Eisenman’s House IV diagrams have an ambiguous quality. These diagrams start from different originating conditions, such as a linear formal element, a surface, or a volume, understood as a state of becoming into a novel critical architectural meaningful sign deconstructing conventional architectural signs. Although sometimes diagrams, representing ideas may become a qualisign,¹⁰ a quality that works as a sign but it is not a sign unless it is embodied, as a diagram can be too abstract from any media or medium, therefore an icon representing linguistic knowledge but without media or medium specific content. 

Eisenman’s House II’s (1968–70) axonometric projection problematizes planar, corner and volume relationships,¹¹ placing in contrast frontal flat Palladian pictorial façade-planes in contrast to corner conditions, developing autonomous formal organizations that create emergent architectural signifiers. The frontal flat planes of Eisenman’s House II’s façades sometimes become a part of a column-sign when planes acquire an embodied thickness and intersect each other forming a corner, an ambiguous double reading enabled by the axonometric projection. Andrea Palladio played critical architecture-painting games with Jacopo Tintoretto,  in which they exchanged media: while Palladio made façades as pictorial painting planes flattening-out space, Tintoretto displaced the real planarity of painting by motivating depth through illusionistic perspective. In the case of Palladio, Hejduk, and Eisenman, although there is a translation of content between media from painting to architecture, there is a media specificity means to activate meaningful signs which sometimes may become signifiers. 

We identified visual semiotic signs becoming signifiers activating feedback within a Representation Restricted Reality” (RRR).¹² We defined a RRR in which we can only access reality through limited, incomplete ¹³ semiotic systems of representations: from mathematics’ arithmetic symbols to visual systems of representation, such as perspective or axonometric projection; to computational algebraic symbols and binary processing signals. When we represent reality through systems of representation, we build up a separation from reality: from perspective’s vanishing point to axonometric projections that help measure and control space as an object. Reality is both measured, understood, and processed through our senses in eye-mind interaction, reconstructing reality in relation to the model-framework implemented, establishing a bias model that enhances certain aspects of reality against others unknown to the human mind. Moreover, our brain neuronal structure models itself constantly after our understanding of reality. 

Some neuroscience research claims that the grammar of our natural language influences the structure of our brain’s neural connections since different languages activate different parts of our brain activating a neuroplasticity,¹⁴ therefore we defined language as a form of program. Dealing with this struggle can help us know more how we understand, perceive, measure, process, and replicate what we consider reality and nature to at least explore our own limited spectrum and how our mind works. This process is also a feedback process, therefore restricted to loops, since we not only understand reality through symbolic form but reality itself becomes modeled and structured physically through the ideology of the semiotic representation systems implemented. Reality is restricted to systems of representation, therefore we argue that we live within a RRR, since these systems serve as responsive frameworks to expand our brain neuronal structures as they expand reality itself modeled after the systems that measure the same reality. 

Perhaps due to a positivist technocratic validation system, in which technology is functional knowledge, computer scientists as well as programmers, or simply some of the current generations, are less aware of problems of representation. In complex layered AI frameworks, conventional systems of representation, such as perspective, become normalized as they are used by programmers. In fact, they project cultural conventions without understanding the necessary deconstruction that is needed to advance a more progressive agenda of architecture. Such process needs to be conducted at a deeper level, which is compatible with a more updated understanding of reality. This is where we try to politically operate activating an architecture of information compatible with an updated critical understanding of reality. We argue that if we do not challenge conventions in systems of representation, we cannot claim to be architects-authors. Computation is anticipatory, the same way that systems of representation structure architecture and language anticipates our thoughts. 

We must challenge the conventional systems of representation and digital signifiers that are implicit in complex neural networks layers in artificial intelligence, otherwise we cannot claim we are authors of our architecture. Meanwhile we are not only letting Big Tech plagiarize our work but we are helping train their repositories through reinforced learning through human feedback (RLHF). Representational problems as discussed in Charitonidou’s book and in this review attempt to help us challenge the systems of representation implicit today, particularly to be able to understand architecture through computation.