Since the late 1980s, technology-driven practices have transformed how architects generate and visualise designs, with technology primarily serving as a tool for advanced calculation, drafting, and modelling. Recent advancements in computer vision and deep learning, however, have introduced more sophisticated methods for analysing, retrieving, and generating design representations through enhanced visual recognition and reasoning capabilities. This evolution has expanded the functionality of design tools beyond conventional drafting and modelling into realms of conceptualization and brainstorming, processes that rely on various forms of intelligence. 

Intelligence plays a critical role in the iterative process of creating, interpreting, and refining architectural representations. This cycle, transforming architectural ideas from abstract concepts into concrete spatial forms, increasingly involves collaboration between human and machine intelligence. Reflecting on these evolving practices that integrate human and machine intelligence in design requires a deeper understanding of the forms of intelligence involved. 

Intelligence is inherently multifaceted (Gardner, 2011), making it challenging to distinguish which types of intelligence are engaged during the design process (Cross, 2010). These include visual intelligence (Arnheim, 1969; Goldschmidt, 1994), spatial and linguistic intelligence (Gardner, 2011), analytical, creative, and practical intelligence (Sternberg, 1985), fluid and crystalized intelligence (Cattell, 1963), distributed intelligence (Salomon, 1993), collective intelligence (Lévy, 1997), and artificial intelligence (Turing, 1950; Russell and Norvig, 2016). By examining these types of intelligence within architectural representation cycles, the study aims to elucidate the dynamic interplay between human and machine cognition in shaping contemporary design theory and practice.

Central to this study is the 'architectural representation cycle,' a conceptual model defining four stages of design representation: experiencing the building spatially, envisioning it mentally, describing it visually or linguistically, and encoding it digitally. This loop highlights the transition from internal concepts to externalised forms—whether pictorial, linguistic, or symbolic—each integral to translating abstract spatial ideas into building forms.

Mental representation refers to the architect's internal construct of a building, which serves as a cognitive framework guiding the design process (Goldschmidt, 1995). Such mental constructs enable architects to understand and navigate complex information. Internal mental representations are frequently externalised into pictorial forms—both analogue and digital—such as sketches, drawings, and renders that visually communicate spatial ideas. Pictorial representations play a crucial role in design (Do and Gross, 1996) by functioning as formal systems that facilitate comprehension and navigation of typically ill-defined architectural problems (Koutamanis, Timmermans and Vermeulen, 1995). Linguistic representations, on the other hand, provide precise articulation of design intent, offer contextual explanations, and support effective communication within multidisciplinary teams. Finally, symbolic representation encodes design concepts in formats that computers can store, transform, and analyse, thereby augmenting human capabilities through digital memory and computational processing power (Koutamanis, 2019).

Together, these diverse forms of representation allow architects to transition from abstract spatial ideas to technical drawings, detailed models, immersive renders, and essential project documentation such as textual reports and data tables. Each type of representation has distinct strengths: pictorial representations resemble real objects, making them intuitive for human interpretation, while symbolic. computational representations are precise and suited to mathematical formalisation and machine processing. Pictorial forms allow for creative ambiguity, whereas symbolic forms enable high precision. The choice of representation depends on the task at hand, available resources, and the preferences and capabilities of the ‘observer’—whether human or machine—in interpreting information when presented in a specific form.

To uncover the intelligence embedded within the architectural representation cycle, which involves both human and machine actors, this study draws insights from fields actively defining intelligence, such as psychology, cognitive science, and computer science. The aim is to connect and map specific forms of intelligence within the architectural representation cycle scheme, deepening understanding of how these forms interact and contribute to the design process. By examining architectural intelligence through the lens of AI technologies, this research seeks to enhance the conceptual understanding of artificial intelligence in design, highlighting the growing interdependence between digital tools and design thinking that will shape the future of the built environment.

This text is based on a presentation delivered at the 18th SIG Design Theory Workshop (Paris, 2025) and was subsequently developed into a guest lecture for an Intelligence course at ETH Zurich run by Prof. Dr. Christoph Hölscher (Chair of Cognitive Science). 

References

• Arnheim, R. (1969) Visual thinking. Berkeley: University of California Press.

• Cattell, R. B. (1963) Theory of fluid and crystallized intelligence: A critical experiment. Journal of Educational Psychology, 54(1), 1–22 

• Cross, N. (2010) Design Thinking as a Form of Intelligence. In: Interpreting Design Thinking (Dorst, Kees and Stewart, Susan eds.), Design Thinking Research Symposia, Design, Architecture & Building, University of Technology Sydney, Sydney, NSW, Australia, pp. 99–106. 

• Do, E.Y.L. and Gross, M.D., 1996, June. Drawing as a means to design reasoning. In AI and Design. Stanford, CA, USA: Stanford University.

• Gardner, H.E., 2011. Frames of mind: The theory of multiple intelligences. Basic books.

• Goldschmidt, G., 1994. On visual design thinking: the vis kids of architecture. Design studies, 15(2), pp.158-174.

• Goldschmidt, G., 1995. Visual displays for design: Imagery, analogy and databases of visual images. Visual databases in architecture, 1, pp.53-74.

• Koutamanis, A., Timmermans, H.J.P. and Vermeulen, I. eds., 1995. Visual databases in architecture: recent advances in design and decision making. Avebury

• Koutamanis, A., 2019. Building Information-Representation and Management: Fundamentals and Principles. TU Delft OPEN Publishing.

• Lévy, P. (1997) Collective Intelligence: Mankind’s Emerging World in Cyberspace. Cambridge, MA: Perseus Books. 

• Salomon G. (Ed.). (1993). Distributed Cognitions: Psychological and Educational Considerations. Cambridge: Cambridge University Press. 

• Sternberg R.J. (ed.) The Cambridge Handbook of Intelligence. Cambridge: Cambridge University Press (Cambridge Handbooks in Psychology), pp. 780–801

• Russell, S. and Norvig, P. (2016) Artificial intelligence: A modern approach. 3rd ed. Upper Saddle River, NJ: Pearson.

• Sternberg, R. J. (1985) Beyond IQ: A triarchic theory of human intelligence. Cambridge University Press.

• Turing, A. M. (1950) ‘Computing machinery and intelligence’, Mind, 59(236), pp. 433-460.