Artificial intelligence assessments of façade similarity: Comparing AI assessments using Christopher Alexander’s 15 properties of living structure, and human evaluations

Aesthetic judgement in architecture—a capacity to differentiate building façades and forms—is crucial to many built environment policies, processes and determinations. Until recently, this cognitive skill has been the sole domain of humans, whether laypeople or experts, who make intuitive or informed assessments of the visual properties of buildings. However, increasingly, Artificial Intelligence (AI) is being used for this purpose, as it appears to be able to replicate simple human aesthetic reasoning tasks, such as determining how similar or different two façades are. There has been a parallel rise in AI-enabled approaches optimised in response to various architectural theories. Several recent examples use Christopher Alexander’s famous theory of ‘living structure’ and its 15 properties to drive AI-enabled façade assessments. Despite the analytical and regulatory potential of these approaches, neither has been extensively tested against human data to assess its validity. In response, this research reports the results of the first study to compare: (i) baseline AI assessments of façade similarity ( n = 204 façades), (ii) AI assessment using Alexander’s 15 properties for living structure of the same 204 façades, and (iii) human similarity assessments of these façades ( n = 147 participants). To address the knowledge gaps, two pairs of hypotheses are framed for testing using statistical analysis. The research finds that AI can broadly approximate human perceptions of similarity, but they are inconsistent and require nuanced interpretation. Nevertheless, the research demonstrates sufficient utility to suggest that AI could be refined to better match human perceptions, thereby providing a valuable method for professional design and planning assessment.