Simulation does not belong exclusively to digital culture; it has deep historical roots in analog modeling. “At every stage of technique since Daedalus or Hero of Alexandria,” writes Norbert Wiener, “the ability of the artificer to produce a working simulacrum of a living organism has always intrigued people.”1 The limits of modeling—of abstraction, reduction, formalization—foreground that they are self-referential structures. This is apparent in the computational practices commonly labeled as simulation today, but it is worth teasing out some uncanny effects of their Möbius-strip character: we expect experts to be trained on simulations, and simulations in turn to be trained by experts. When cybernetic models exhibit their own dynamic instead of erasing their constitutive contributions to knowledge production, they resist becoming imperceptible (or anesthetic) and appear as media.
In The Architecture of the Uncanny: Essays in the Modern Unhomely (1992), Anthony Vidler draws on this idea to read architecture as wound up in repetition, doubling, and existential strangeness.2 Yet his titular word “unhomely” cannot be found in any dictionary: as transliteration of a German word, it tries an alternate translation for unheimlich—perhaps to avoid the uncanny repetition of the word uncanny. “Homely,” however, has four entries in Webster’s Ninth New Collegiate Dictionary: “1: suggestive or characteristic of a home; 2: being something familiar with which one is at home; 3a: unaffectedly natural: SIMPLE; 3b: not elaborate or complex; 4: plain or unattractive in appearance—homeliness.”3 What would its inverse be: not suggestive of or uncharacteristic of a home, unfamiliar, affectedly unnatural, elaborate, complex, and attractive in appearance? If the uncanny, as Vidler writes, “not unnaturally found its metaphorical home in architecture,” this need not mean only buildings: writings and drawings, as part of architectural practice, are also folded in and doubled over.
It might be an overstatement to suggest that the practice of architecture pivots around simulation, much as one sees models at its core. Instead, it is worth considering whether architecture as a discipline could be said to revolve around testing the epistemic status of models. For there is an argument to be made that architecture always already operates both as a practice and theory of models, models of environments and models for environments. A model of something helps render a question to make it tractable and accessible, whereas a model for something implies a guiding or normative role.4 Both are exercises of the imagination: interactions with a model can fulfill expectations and confirm theoretical hypotheses, or violate them and thus question the basis for modeling. According to Willard McCarty, “as a tool of research … modeling succeeds intellectually when it results in failure.”5 The feedback loops this creates can be described in different ways, but it is insufficient to describe architecture as either descriptive—offering a taxonomy of buildings and methods—or prescriptive—in contextual regulations or in the preparation of plans.
Testing Reality
Arguably, any process of representing the dynamic behavior of one system in the behavior of another system is not anti-referential; simulation techniques aim to faithfully reproduce selected characteristics of something that remains a referent. There is always a risk that abstraction or approximation makes models lose precision or omit variables, or inversely that the rich data of a simulation model make us forget about assumptions baked into the model. The use of models has a long tradition in education, from before Francis Bacon’s New Atlantis (1626) provided a “model for a college” to after Descartes invited his readers to emulate his method. The term has often been taken literally, as when artists including Leonardo, Michelangelo, Dürer, or Galileo, as well as the philosopher Leibniz, drew and built fortification models. Closer to our time, the MIT Tech Model Railroad Club (a train set maintained in the 1950s and ’60s with an elaborate communication system) gave rise to the hacker scene there, and there is a long tradition of the model airplane, from before the Link Trainer to after the current generation of Microsoft Flight Simulator.
So let us not indulge in a totalization of simulation that undermines any distinction between original and copy, as a number of anti-Platonist thinkers had it. Baudrillard saw a classical age of counterfeit (interfering in the distinction between reality and appearance) giving way to an industrial age of communication media (interfering in the distinction between production and reproduction) and on a trajectory into an age of simulation based on the manipulation of code to model and test, to encrypt and decrypt. Without suggesting we go along with his hyperbole, Baudrillard is not wrong to alert us to the rise of models, games, tests, tree diagrams, flow charts, and other digital operational forms that produce a new mediality. “After the metaphysic of being and appearance, after that of energy and determination, comes that of indeterminacy and code,” he surmised, characterized by “cybernetic control, generation from model, differential modulation, feed-back.”6 The transition he anticipated “from a capitalist-productivist society to a neo-capitalist cybernetic order,” giving rise to “social control by anticipation, simulation, and programming” meant that simulations uncannily test reality rather than the other way around, and it is this figure of uncanny inversion I wish to focus on here.7
Since the 1950s, scholars foresaw that once digital computing became fast enough, it would furnish “simulation for vividness” in models that strike observers as clear and convincing: “simulation for deduction and exploration” making questions tractable in new dimensions, and “simulation as archive,” whereby models store the collected knowledge of an entire discipline.8 Data-driven models help predict processes such as climate, weather, traffic, car accidents, or explosions: comparison of simulation results with observed data leads to a more accurate model by tuning corresponding parameters. Existence-proof models seek to prove the hypothetical existence of phenomena that have not yet been observed or explored in certain parameter ranges. The goal of exploratory models is to study problems for which a theory is not (yet) available in computer programs that function like thought experiments. When machine learning was formulated as a field of inquiry in the 1950s, it pitched an agenda of simulating “every aspect of learning or any other feature of intelligence” in machines.9
Simulation in Architecture
Once computing began to support interactive models, architecture began exploring their transformative potential, as when Gordon Pask suggested that one can deploy simulations to modify and improve interactions between designers and systems they designed: “Once architects understood themselves as systems architects, he suggested, they could cybernetically change their design practices.”10 One simulator instantiating this idea is the ARCHITRAINER developed by Chris Abel at the Architecture Machine Group at MIT in the 1970s as a training model of and for architectural practice, using “computer-aided instruction (CAI) to simulate ‘dialogues’ between student architects and hypothetical clients.”11 As a simulation of the relationship between architects and clients, ARCHITRAINER modeled interactions on three levels: at the tutorial level, artifacts and client preferences were modeled; at the level of gamification, students explored that model; and at the control level, the system would assess the student’s understanding of the clients’ requirements.12
Abel’s interactive computer game simulates multiloop feedback systems, explicitly leaning on cybernetics research by Ross Ashby, Stafford Beer, Norbert Wiener, and Gordon Pask.13 Pask argued that before the early nineteenth century, architecture was dominated by conventional rules and rigid codes, but designing a railway station or an exhibition hall were projects that could not simply be executed with reference to the old rules. As architects took an interest in organizational systems, they discovered communication and control—but without a fully developed conceptual framework.14 If the “abstract concepts of cybernetics can be interpreted in architectural terms (and, where appropriate, identified with real architectural systems),” this leads to a cybernetic theory of architecture.15
Now that artificial intelligence is the latest guise of the uncanny, consider how much simulation has become part and parcel of modeling and planning disciplines, including architecture. We are surrounded by AI illustrations of layouts or interiors, VR walk-throughs of architectural designs, 3D drawings used in construction or approval processes, or other computational techniques: their increased realism lowers our tolerance for abnormalities and imperfections, which feel uncanny. One of the most hyperbolic formulations of architecture as simulation that becomes indistinguishable from what it models is Ivan Sutherland’s provocative thought experiment of the ultimate display, a vision that spurred a lot of investment in augmented reality and virtual reality technologies. In an infamous 1965 lecture, he suggested a hyperreal interface: “a room within which the computer can control the existence of matter. A chair displayed in such a room would be good enough to sit in. Handcuffs displayed in such a room would be confining, and a bullet displayed in such a room would be fatal. With appropriate programming such a display could literally be the Wonderland into which Alice walked.”16
As we are surrounded by VR and AI hype today, have we come to understand consciousness merely as information processed in certain ways, regardless of what machine or organ is used to perform that task? Or is this invalidated by reminding ourselves that a hurricane model does not make anything wet, a fusion plant simulator does not produce energy, a metabolic model does not consume actual nutrients—so a model of the human brain does not amount to giving rise to thought? As digital technology affords fascinating modes of time-axis manipulation and detailed rendering, increasingly sophisticated simulations can model selected attributes in such a way as to allow them to be inverted along their axes, their envelopes turned inside out. This is useful for practical purposes and highly productive in conceptual approaches, but it remains uncanny. We might do well to remind ourselves of the fundamental difference between fiction and simulation: the former cannot be verified or falsified, while the latter is a science of the artificial that requires continuous critical testing.17
Norbert Wiener, Cybernetics (Cambridge: MIT Press 1985), 39.
Anthony Vidler, The Architectural Uncanny: Essays in the Modern Unhomely (Cambridge: MIT Press 1992), 13.
Webster’s Ninth New Collegiate Dictionary (Merriam Webster’s Inc., 1990), 577.
Eric Winsberg, “Sanctioning Models: The Epistemology of Simulation,” Science in Context 12, no. 2 (1999): 275–92; Paul Roth, “Simulation,” in Encyclopedia of Computer Science, ed. Anthony Ralston (New York: Van Nostrand 1992), 1204.
Willard McCarty, “Modeling: A Study in Words and Meanings,” in A Companion to Digital Humanities, eds. Susan Schreibman et al. (Malden: Blackwell, 2004), 256.
Jean Baudrillard, Simulations, trans. Paul Foss et al. (New York: Semiotext(e), 1983), 103.
Baudrillard, Simulations, 111; Andreas Huyssen, “In the Shadow of McLuhan: Jean Baudrillard’s Theory of Simulation,” Assemblage 10 (1989): 6–17; Brian Massumi, “Realer Than Real: The Simulacrum According to Deleuze and Guattari,” Copyright 1 (1987): 90–97; Manuel de Landa, “Virtual Environments and the Emergence of Synthetic Reason,” in Flame Wars, ed. Mark Dery (Durham, NC: Duke University Press 1994), 793–815; Slavoj Žižek, “Cyberspace, or the Unbearable Closure of Being,” in The Plague of Fantasies (London: Verso 1997), 127–67.
Peter Krapp, Computing Legacies: Digital Cultures of Simulation (Cambridge, MA: MIT Press 2024).
John McCarthy et al., “A Proposal for the Dartmouth Summer Project on Artificial Intelligence,” 1955, ➝.
Molly Wright Steenson, Architectural Intelligence: How Designers & Architects Created the Digital Landscape (Cambridge, MA: MIT Press 2017), 17.
Chris Abel, Architecture & Identity (Oxford: Architectural Press 2000), 33. The architecture machine group at MIT was a predecessor to the Media Lab, and its cofounder Negroponte wrote his master’s thesis on simulation and urban environments. See Nicholas Negroponte, “The Computer Simulation of Perception during Motion in the Urban Environment” (master’s thesis, MIT, 1966), submitted to a committee including architects Gyorgy Kepes, Aaron Fleisher, Wren McMains, Imre Halasz, and Leon Groisser; and Stewart Brand, The Media Lab: Inventing the Future at MIT (New York: Viking, 1987).
Chris Abel, “Analogical Models in Architecture and Urban Design,” METU Journal of Faculty of Architecture 8, no. 2 (1988): 175–76; Audrey Watters, Teaching Machines: The History of Personalized Learning (Cambridge, MA: MIT Press 2021); Gordon Pask, “Machines That Teach,” New Scientist 234, May 11, 1961, 308–11.
John Hamilton Frazer, “The Cybernetics of Architecture: A Tribute to the Contribution of Gordon Pask,” Kybernetes 30, nos. 5–6 (2001): 641–51.
Ensar Temizel, “The Cybernetic Relevance of Architecture: An Essay on Gordon Pask’s Discourse on Architecture,” Proceedings of the 38th eCAADe Conference on Education and Research in Computer Aided Architectural Design in Europe, eds. Liss C. Werner and Dietmar Koering (Hamburg: Tredition, 2020), 471–80.
Gordon Pask, “The Architectural Relevance of Cybernetics,” Architectural Design 37, no. 6 (1969): 494–96; Nicholas Negroponte, Soft Architecture Machines (Cambridge, MA: The MIT Press, 1975).
Ivan Sutherland, “The Ultimate Display,” Proceedings of IFIP Congress (1965), 506–8.
Herbert Simon, The Sciences of the Artificial (Cambridge, MA: MIT Press 1998), 4.
