AI-driven human-in-the-loop structural optimization of gridshells

Recent advances in AI have improved automation and optimization capabilities. However, current systems remain limited in reproducing the intuitive and subjective aspects of human creativity. In architectural and structural design, this creates a gap between performance-driven optimization methods and the designer’s qualitative intent, as existing workflows rarely integrate human preferences within a rigorous optimization process. This paper introduces a Human-in-the-Loop optimization framework that incorporates designer preferences as part of a multi-objective formulation. The method combines parametric modeling, evolutionary optimization and machine learning to guide the search toward solutions that balance efficiency, sustainability and creative intent. The framework is validated through a user study in which participants generate distinct Pareto fronts aligned with individual preferences. Results demonstrate that the approach effectively embeds human feedback within structural optimization, fostering more creative and informed exploration in preliminary design.