Autonomous Design Agency in AI-Assisted Architecture: Exploring Intentionality, Ethical Autonomy, and Design Innovation Through Philosophical Pragmatism

Abstract

This study reformulates autonomous design agency in AI-assisted architecture as a measurable engineering problem rather than a purely philosophical one. A pragmatist control framework was developed to evaluate how bounded AI autonomy affects design quality, regulatory compliance, and stakeholder acceptance in three completed projects: a sustainable office building, a community park, and an urban renewal district. The dataset combined archived design logs from 35 iterations, site measurement records, interviews with 15 designers, and end-user evaluations from 120 respondents. The method operationalized functional intentionality through goal-directed performance improvement, using an Autonomy Allocation Index, a weighted composite performance score, and project-specific environmental and operational indicators. The results show that higher bounded autonomy was strongly associated with better overall performance (r = 0.921, p < 0.001). In the office case, energy use intensity decreased by 20.3% while daylight factor increased by 59.4%; in the park case, biodiversity increased by 71.9%; and in the renewal case, function integration improved by 58.5% while zoning compliance reached 98%. User satisfaction differed significantly among the three outcomes (F(2,117) = 6.041, p = 0.003). The study provides a reproducible engineering framework for calibrating AI autonomy in architectural workflows without weakening human responsibility.