Preference assessment of generative AI image tools for architectural design generation using MS-TORO framework

Since the emergence of artificial intelligence (AI) tools in manufacturing and service industries, their potential has also been recognized in more immersive and personalized domains, such as architectural design development. While generative AI (GenAI) image tools offer valuable support in creating architectural concepts, their capabilities in areas such as usability, user experience, emotional resonance, and user satisfaction still require enhancement to support full integration into design workflows. Addressing this gap, this paper introduces the novel application of a multiple-stakeholder target-oriented robust optimization (MS-TORO) framework to evaluate and improve the performance of GenAI image tools within the architectural domain. This is the first study to apply MS-TORO in this context, enabling a structured, data-driven approach to balance competing design criteria across diverse stakeholder perspectives. Among the tools assessed, Adobe Firefly 2 emerged as the most suitable, contingent on specific design adjustments to meet stakeholder-defined targets.