Large language models (LLMs) judge three pairs of architectural design proposals which have been independently surveyed by opinion polls: department store buildings, sports stadia, and viaducts. A tailored prompt instructs the LLM to use specific emotional and geometrical criteria for separate evaluations of image pairs. Those independent evaluations agree with each other. In addition, a streamlined evaluation using a single descriptor “friendliness” yields the same results while offering a rapid screening measure. In all cases, the LLM consistently selects the more human-centric design, and the results align closely with independently conducted public opinion poll surveys. This agreement is significant in improving designs based upon human-centered principles. AI helps to illustrate the correlational effect: living geometry → positive-valence emotions → public preference. The AI-based model therefore provides empirical evidence for a deep biological link between geometric structure and human emotion that warrants further investigation. The convergence of AI judgments, neuroscience, and public sentiment highlights the diagnostic power of criteria-driven evaluations. With intelligent prompt engineering, LLM technology offers objective, reproducible architectural assessments capable of supporting design approval and policy decisions. A low-cost tool for pre-occupancy evaluation unifies scientific evidence with public preference and can inform urban planning to promote a more human-centered built environment.
Smith et al. (2025) studied this question.