Artificial Intelligence and Architectural Empathy

Abstract

Artificial intelligence (AI) is transforming numerous disciplines, including architecture, at an increasingly rapid pace, making it difficult to grasp the full implications and consequences. Although terms such as “intelligence” and “learning” may give AI a human-like appearance, they cannot hide the fact that AI is not, and will never be, human. In this paper, we propose a thought experiment in which we temporarily assume the quasi-human nature of AI to explore how architectural knowledge might be generated and interpreted by such systems. Using a series of verbal and visual prompts, we examine the extent to which AI can demonstrate what we refer to as “architectural empathy”—a metaphorical concept that probes AI’s capacity to engage with the experiential and spatial dimensions of architecture. AI’s limitations become visible not when architecture is treated as an object of description, but when it is treated as an experience of inhabitation. Understanding how this architectural empathy is structured is fundamental, as more people will develop their own understanding of architecture through AI and tools such as ChatGPT.