Academic Awards 2025 booklet

47 Comparing Human, Human-AI, and AI-Automated Structural Design Processes and Products for Zoning and Stabilization Building design has predominantly remained a human task, being labor-intensive and prone to human error. Hence, this thesis investigates the effect of integrating Artificial Intelligence (AI) in the human structural design process (Figure 1), aiming to benefit both design teams and the outcomes of their work. An experiment is developed involving groups of students stabilizing structures (i.e. preventing kinematic mechanisms) either manually, fully via AI, or through human-AI collaboration (Figure 2a). We extend the experiment with a second assignment involving ‘zoning’ (Figure 2b), which uses pattern recognition to find optimal spatial layouts that allow for efficient structural systems. To enable interaction between a human and AI agent, the needed tooling has been developed. The students’ processes are captured through a linkography analysis (protocol analysis), thus enabling quantitative comparison. Even though the results show that AI agents outperform participants regarding task-specific products and the total time spent, human-AI collaboration shows most potential. The AI agents streamline human processes (regarding steps, efficiency, complexity) without diminishing creative freedom (thinking out-of- the-box). Nevertheless, collaboration with AI agents enhances human products and processes only when participants define a clear problem-solving strategy before asking for AI suggestions. Solely relying on AI from the start proved counterproductive. Figure 2: A student adding elements to the initial structure with the aim of stabilizing it in collaboration with AI suggestions (a), and one performing a zoning assignment with a corresponding information sheet, currently modifying the Building Spatial Design (b). Figure 1: Comparison of three distinct design approaches, each characterized by a different level of AI involvement: human design, human-AI design, and AI design.