Emergent Behaviors in Generative Artificial Intelligence in Design

Richa Gupta examines the transmission of social bias through generative artificial intelligence and its consequences for design, particularly within the built environment. Through examples drawn from architecture and text-to-image systems, she demonstrates how historically embedded assumptions within data are reproduced in AI outputs and come to shape early design decisions whose effects scale across buildings, cities, and infrastructures. The talk critiques the prevailing use of generative models as instruments of visual production divorced from material, environmental, and economic constraints. In response, Gupta articulates a framework for integrating domain-specific, real-world data into generative workflows, situating AI as a contextual and deliberative tool that supports responsible design judgment rather than substituting for it.