PREDICTIVEDESIGN: AI, NEUROSCIENCE, AND ARCHITECTURE Book Chapter

Leach, N. (2026). PREDICTIVEDESIGN: AI, NEUROSCIENCE, AND ARCHITECTURE . 135-159. 10.4324/9781003347330-12

cited authors

  • Leach, N

abstract

  • Neuroscientists have long argued that human perception – and, by extension, intelligence – is based on prediction. Meanwhile AI itself is also inherently predictive. This is how ChatGPT works, by predicting the next token or word. Moreover, prediction, it could be argued, also extends into architectural design. Architects, like neural networks, are trained to perceive the world through predictive filters, ‘hallucinating’ structures into architectural forms. Inspiration often arises from this predictive bias, as in Utzon’s Opera House, reimagining sails as vaults. AI image generators, such as Midjourney, reflect similar processes, combining previous concepts to produce new designs. The parallel suggests that architectural creativity, far from ineffable genius, is a predictive recombination of experience and training. Meanwhile, as we move into the world of spatial intelligence and agentic AI, we are beginning to glimpse the emergence of a new form of predictive design. Here, architecture is no longer defined by form, representation, or static objects. It is defined by anticipation. The era of ‘predictive design’ has arrived.

publication date

  • January 1, 2026

Digital Object Identifier (DOI)

start page

  • 135

end page

  • 159