Between standardized bodies and spatial form, architectural data has long been grounded in measurement. Manuals such as Architects' Data by Ernst Neufert, and the anthropometric studies of Henry Dreyfuss, translated human dimensions, movements, and behaviors into design parameters. Here, data operates as a calibrated interface between the human body and the built environment, encoding reach, clearance, posture, and use into standards that shape space. 

However, if these standards frame data as stable, measurable, and human-centered, contemporary computational systems begin to unsettle this view. Data is no longer limited to fixed bodily dimensions or predefined use, but expands into dynamic and heterogeneous forms—an active medium through which relationships, patterns, and meanings can be extracted and reconfigured. 

In the computational context, any digitally recorded representation of information that can be stored, processed and transmitted electronically is considered data, including various electrical signals, text, images, spreadsheets, videos, audio files, and more. Digitally stored data can be thought of as a record of information that can be retrieved from memory at a later time or computationally processed for a specific purpose. Different types of data require different computational techniques to be analysed effectively. For decades, processing numbers and signals has been the primary focus of computational processing, but with the advent of deep learning artificial intelligence technologies, it is now also possible to process other forms of data, such as image (visual) and text data.

From an artificial intelligence perspective architectural design data can be defined as all digital content generated and transformed in the design, planning, construction, and building operation process – from drawings, renders and maps (‘visual data’), to memos, reports and design briefs (‘textual data’), to numerous models, schedules and parameters (‘numerical’, or ‘tabular’ data). Each of these data types requires different computational techniques to be analysed and is linked to a specific AI field of expertise. Computer Vision focuses on ‘visual data’, Natural Language Processing on ‘textual data’, and Machine Learning on ‘numerical data’. It is worth noting that all of these different types of data are semantically interconnected: from a design brief in which ambitions and design requirements are described through text, numbers and flow charts; to models and floorplans spatially framing the design ambition; to various tables (schedules) describing characteristics of various building components; to realistic visualisations simulating the envisioned spatial situation; and finally to detailed drawings and written specifications describing how to construct the building in the real world.

Moreover, the sequence of design data does not end with the finalization of the design process, but it continues to grow over time. Once a building is completed, its life begins. People take pictures of the building with their smartphones, drones, and satellites, capturing activities that occur inside and outside of it. Through various sensors and recording devices people collect information about the building's indoor environment, measuring temperature, humidity, light, and occupancy to assess and improve energy usage and building operations. The interconnectedness of information across different modalities of data suggests a rich embedding of design knowledge captured in the digital memory that could be harnessed through AI technologies.

By observing and analysing architectural memory records in large quantities, AI technologies could enable architects to compare design information across multiple projects, recognize design patterns, understand casual relationships, spot exceptions and outliers, and much more. As the volume of digital data generated by architects and buildings continues to grow over time, it is becoming more critical to empower computers to read, transform, and analyse architectural digital memory records within and across projects, to support design development, improve design workflows and uncover new insights into building design, performance and human activity patterns. This is especially important considering the growing complexity of design assignments resulting from the challenging environmental and societal circumstances, and the expanding collaborative practices introducing more specialised knowledge in the development of projects.

This text is based on a presentation delivered at the co-creation workshop 'Machine Learning Data for Architectural Design', as part of the CAAD Futures 2023 Conference (Delft, 2023), and was subsequently developed into a lecture for the MSc2 Studio 'Inhabiting Data' at TU Delft. 

References

• Dreyfuss, H. (1960). The measure of man: Human factors in design. See also: Dreyfuss, H. (2002) The measure of man and woman: Human factors in design. Revised ed. New York: Wiley.

• Manovich, L. (2020) Cultural analytics. Cambridge, MA: MIT Press. — his definition of data is used in the text

• Neufert, E. (2012) Architects’ data. 4th ed. Oxford: Wiley-Blackwell.

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