摘要
This paper introduces the "From Design to Dwelling" framework, an HCI-driven approach to bridging architectural design intentions and lived experiences. Central to this framework is the concept of Human-Building Foundation Model (HBFM), which synthesizes multimodal empirical data, including visual, linguistic, and embodied dimensions-to inform AI-aided design tools optimized for human well-being. Our preliminary study analyzed patterns from human annotations, machine-generated visual features, and neural data during first-person navigation of virtual residential interiors across various design styles. Results demonstrate a double dissociation in neural processing: the parahippocampal place area (PPA) correlates with objective visual features, while the retrosplenial complex (RSC) aligns with subjective spatial experiences. While the HBFM is currently conceptual, this work provides an empirical foundation for its future implementation. This work advances both HCI and architectural design by providing empirically grounded methodologies for developing user-centered tools that bridge expert knowledge with quantifiable human experience within architectural spaces.