Computational Tendencies

The Analytical Engine has no pretensions whatever to originate anything. It can do whatever we know how to order it to perform. It can follow analysis; but it has no power of anticipating any analytical relations or truths. Its province is to assist us in making available what we are already acquainted with.
—Ada Lovelace1

Users of Autodesk AutoCAD worldwide are estimated to be 19.3 million.2 Robert McNeel & Associates Rhinoceros has (at least) 1 million users,3 and the visual programming environment plug-in Grasshopper (at least) 130,000.4 Most contemporary architectural production happens within these software, and thanks to the algorithms behind their most simple tools. Scaling, mirroring, and other geometrical functions, and their potential automation, are the basis of how architecture is drawn and conceptualized today. It could therefore be argued that most, if not all architecture is digital. Yet, what scholars and practitioners alike have called digital architecture over the past decades has been treated only as a sub-genre of architectural discourse as a whole.5 This is largely due to the fact that the theoretical and design work focused on coming to terms with advancements in digital technologies has been more focused in devising a style of design which formalizes computational processes rather than making the case for how digital technologies can help improve the quality of the human habitat.6

While universities around the globe become increasingly populated with unrealizable computational formal experiments, cities worldwide continue to develop with standardized and post-rationalized buildings, where the architect’s agency is in most cases limited to façade treatments. Experiments with digital technologies and automation in architecture therefore remain firmly within academic phantasmagorias. In order to determine what it might mean for architects to speculate on a “good” use of digital technology in design and building, we should turn away from a meta-discourse on computational theory forcefully molded into buildings,7 and instead shift our attention toward re-examining existing and established design and building methodologies within a computational framework. One place to start might be in the connection between artificial intelligence as a set of technologies that externalize certain rational and logical qualities of human reasoning, an early precursor or which can be seen in the tradition of Rationalism within architecture discourse.

The so-called “digital turns” in architecture and construction, which correspond to consecutive waves of integration of digital technologies into practice, have yet to fully comprehend and render productive the link between formal logic, computation, and architecture as a habitat-making practice. While such developments have and continue to promise a new design methodology—by automating specific design processes and driving the development of automation in construction, thereby defining the new role potentially occupied by the architect in the process—there are a series of problems with making the case for automation in architecture today. Firstly, the most recognizable applications of automation in built architecture have manifested as plasterboard and titanium hellscapes (e.g. the work of Zaha Hadid Architects and Gehry Partners), which appear to be more concerned with establishing a “style” and a “brand,” thus reifying computational processes and logics, than with delivering good, well-functioning buildings.

Stanislas Chaillou, Training Sequence, 2019. Source: Medium.

There are already multiple efforts within political theory to establish a different approach to technologies of automation, which look at commoning and other political strategies to critique current applications of augmented intelligence within a predatory neoliberal framework.8 At the same time, many theorists are working on formulating a critique of augmented intelligence technologies to establish what agents are shaping them and how they might control the presence or lack of “sensibility” in how they are designed and used. Franco “Bifo” Berardi, for instance, recently pointed toward the impossible coexistence of sensibility and computation when he decried that “the Two pan-logical projects tower above the history of modernity, the recombinant project of Leibniz and the teleological project of Hegel [erode] the kind of mental activity that cannot be reduced to computation or biological determinism.”9 Connecting computation and sensibility will be necessary to understand how AI can be used for public good.

A major issue with the introduction of automation into construction is that it is difficult to specify what architecture is and does in “computable” terms. Therefore, what is currently being invested by some degrees of automation (i.e. with BIM software) is just the taxonomy of single building components (wall, window, floor). These feed the typological frameworks already established by a real estate market, one predicated on the simplified indexical correspondence between components (or “design features”) and their corresponding value. Within this framework of “simplified automation,” it is impossible to determine any architectural quality beyond these parameters, not to mention impose it as a necessary, generalized concern. Representative of what Charles Babbage envisioned as the division of labor brought about by computational machines,10 the professional boundaries of architecture must be altered to avoid further erasing the specificity of the architect’s contribution to the built environment in a more general paradigm of increased automation.

To claim such specificity, it is necessary to at least try to define what architecture is. For the sake of this argument, let us define architecture as a “holy trinity” of practices. It is the abstract discipline of measuring and seeing the world (in Vitruvius’s words, “the reification of reasoning”), and as such concerned with organizing data, proportions, and harmony. It is the political art of giving spatial definition to lebensformen; of building and giving material form to ways of living individually and collectively. And lastly, it is the management of complex material and productive flows, including the labor implicated in the making of the human habitat. Such a provisional, albeit unambitious definition of what architecture is and does is often missing in the context of “digital architecture,” replaced instead by speculative interpretations of computation and fabrication. This not only leads to the discipline’s further specialization—the consequence of which can be seen in the rise of a myriad of figures that have progressively absorbed specific, and defined, aspects of what was previously encompassed by the professional role of the architect—but also limiting what can be collectively understood as “good” architecture.

The definition of what constitutes “good” architecture is, in fact, always at the center of architecture discourse, despite never finding a definite answer. Discourses around digital architecture have too often resolved the question of the “good” in architecture by escaping into the realm of taste or artistic judgment. This not only obsolesces what the discipline has otherwise understood as “good” architecture, but also condemns its own understanding of the “good” to obsolescence.11

Gregg Lynn FORM, RV Prototype House, 2012, section sketch. Image © Greg Lynn FORM.

This problem was evident even in the so-called first digital turn, when complexity was elevated to architecture’s ideological telos.12 Architects like Greg Lynn were fascinated by how in other disciplines it was increasingly easy to obtain mathematical models of complex natural phenomena at all scales. Complexity thus became formulated as a value in and of itself, one to be sought at all costs. This approach was supported by the work of Gregory Bateson (and also following vestiges of organic architecture theories), who identified symmetry in nature as a loss of complexity, a sort of data saving mechanism.13 Similarly, in Mario Carpo’s formulation, architecture’s raison d’etre is found in the embodiment and reification of advancements in data management, and mathematics and reasoning more generally. The mantra of complexity coupled with the use of emerging technologies derailed the consequent architectural production towards the endless exploration of complex “data intensive” forms. This progressively eroded the capacity of architecture to propose syntheses of the layered and often conflictual conditions that surround its making, and positioned it as a luxury object-making discipline akin to product design. This was justified only by the economic context that the architecture of the first and second digital turns existed within: speculative real estate and its attendant techniques of urban renewal, gentrification, and value production, all of which are predicated on novelty.

The Rationalist tradition in architecture offers an alternative approach to questions of automation. Rationalism was originally formalized during the Enlightenment as an attempt to recast architecture’s epistemological position in a moment when knowledge as such became specialized. Architects of the time sought to root their practice in logical and abstract principles to reinstate architecture as more than the craft of copying existing models from antiquity. Yet in order to do this they studied antiquity. Sebastiano Serlio wrote his Books on Architecture as a manual for architects, in which he extracted and systematized proportional relations from classical buildings into a grammar and syntax, a “system,” which could be applied via “transmutation,” Serlio reduced the available examples into a reliable “code” to serve as a shared language, applicable to any building typology.14 A reduced and precise language of spatial primitives, Serlio’s code can be understood to a proto-BIM system, one whose core values are not market availability or construction efficiency, but harmonic proportions.

Sebastiano Serlio, Book of Architecture, 1537–1575, Book 2, Chapter 3, Fol. 13.

Sebastiano Serlio, Book of Architecture, 1537–1575, Book 2, Chapter 3, Fol. 15.

Sebastiano Serlio, Book of Architecture, 1537–1575, Book 1, Chapter 1, Fol. 7.

Sebastiano Serlio, Book of Architecture, 1537–1575, Book 2, Chapter 3, Fol. 13.

Mosej Ginzburg, the Soviet architect most well-known for his formulation of the “social condenser,” wrote a book titled Rhythm in Architecture in 1922. In it, he analyses the rhythmic qualities of classical architecture and attempts to extract compositional patterns from them. In the opening sentences of the book, Ginzburg states that “the universe is permeated with rhythm,” and goes on to redraw buildings as simple diagrams of their composition, not too distant from Serlio’s Ten Books.

The Rationalist tradition was reappraised in the mid-twentieth century in an attempt to critique the functionalism which had come to characterize modernist architectural production at the time. Starting in 1966, Tendenza came to define a heterogeneous “tendency” in European architecture whose main programmatic objective was the formulation of an “architecture of reason,” one rooted in historical references to classical architecture and architecture from the Enlightenment.15 Of the many texts and projects which attempted to define such methodology, in 1967, thinking through the work of Étienne-Louis Boullée, Aldo Rossi struggled to understand how architecture should be based on logical principles and clear propositions while at the same time be the work of an artist.16 He makes a distinction between a “conventional” rationalism, which cannot resolve the conflict between architecture understood as both a science and an art, and an “exalted” rationalism, which, like the work of Boullee, thrives in this conflict. Rossi understood Boullee’s method as: a synthesis of a logical understanding and ordering of problems; a focus on composition as a way to proportion and measure space; an inspired and emotional relation with nature; and a typological approach rooted in architecture history as a shared language.

In formulating his concept of exalted rationalism, Rossi’s defined architecture as typological practice, one primarily focused on the formalization of emerging uses and needs, as well as compositional methods to proportion space and building parts, therefore referring to both a shared “language” and invention. Today, an architectural methodology which seeks to incorporate automation in design could move in the same direction, finding new ways of organizing space under the increasingly complex circumstances producing it. Composition is not an esoteric art, but a concrete and fundamental attribute of architecture. The capacity to perceive harmony in composition and the role of proportion in space is fundamental. Yet not since the work of Le Corbusier, Hans van der Laan, and Ludwig Wittgenstein has there been a serious discussion surrounding it in architectural discourse.

“The Halfe Ancient Foot,” drawing from Sebastiano Serlio, Book of Architecture, 1537–1575, Book 3, Chapter 4, Fol. 47.

Following these precedents, a third digital turn could be focused on innovating processes and methodologies, looking again at questions of composition, harmony, and intelligibility, and providing new solutions in the field of typology and spatial organization. It could resolve the contradiction between “phenomenologist” and “digital” architects, and be able to deliver material and spatial quality not just at the small scale available to the few, but to the many. Digital architecture today is not one of absolute abstraction, but one that is able to balance the intrinsic abstraction of architecture’s methods and processes and the material and spatial qualities of experience. A third digital turn might actually recognize architecture’s contribution to Rossi’s idea of the built environment as a collective human artwork, and use automation to do it with a spirit of continuity rather than disruption.

In a discussion on automation in architecture, another fundamental question is how to use and where to find data. If the process of designing consists of making creative selections from available information, in the current data-rich context of the built environment, where thinking can be augmented by machines, an expansion of the architect’s technical knowledge is required. If the aim of architectural practice is typological and spatial invention, one might question how a database such as Airbnb’s, which contains precise information on how people inhabit their domestic environment, could be used as a compositional “machine,” not unlike Serlio and Ginzburg’s, to confront clearly defined typological problems in a new paradigm of automated design?

Stanislas Chaillou, GAN-enabled Space Layout under Morphing Footprint, 2019. Source: Medium.

There’s a bright precedent for this idea of data-augmented architecture in the work of Constantinos Doxiadis. The Greek architect, who was employed by the UN to urbanize vast areas of Asia and the Middle East after WWII, was first in Europe to run an architecture office with a calculator.17 This calculator was fed information gathered by the UN in their pioneering data collection and management experiments. If we limit ourselves by looking at the methodological implications of this framework, without entering in the specifics of Doxiadis’s projects, we can see that he was able to inform large scale urban projects with precise data on the social composition of the communities he was designing “for” (some would say against), while maintaining his own specifically modernist methodology, vision, and agenda. In a similar way, we must dare to imagine a use of Big Data beyond its current dystopian implications. Might it not, for instance, help take one step closer to realizing the dream of participatory planning?

Examples of datasets of architectural information from the Rationalist tradition in the seventeenth and eighteenth century include taxonomic typological studies. For example, Le Muet, Briseux and Durand developed treaties and manuals that taxonomized answers to the question of how to build the “most optimal home” (for the standards of the time) in a lot of a given sizes/shapes. Their effort demonstrated a rich array of typological possibilities, all catalogued according to a refined criteria that involved both the programmatic use of space and spatial and architectural quality. Given the abundance of data regarding domestic space, a third digital wave could recuperate these taxonomic studies to be able to, for example, hack programmatic diversity. It could offer a further resolution of responses to rapidly changing needs and uses, which are otherwise regulated either by market forces (which by design reinforce the status quo), or by regulatory frameworks (which are necessarily slow). Algorithmic intelligence might enable us to look at the “housing problem” not just in a speculative manner, but in a systematic way that takes into account the realities and multiple agencies involving and shaping the contemporary domestic landscape. This would not only help design better homes, but also reinsert political agency in the process.