Self-Reference Design: A New Potential Design Methodology in the Age of Artificial Intelligence

Abstract An artificial intelligence (AI)-based design methodology is introduced to explore design through the lens of self-reference. Geometric and design principles are reinterpreted within this framework. Self-referential neural networks, grounded in self-replication, are developed to achieve greater self-referentiality in the AI model. The inherent continuity between self-referentiality in traditional design and the proposed AI-driven design approach is delineated. An experiment is carried out to demonstrate this approach’s efficacy. Ultimately, this methodology is positioned within the discourse of speculative design, envisioning a future where design possesses escalating self-referentiality and autonomy, potentially leading towards “self-inspired” design.