Generative AI is increasingly applied in architectural research, from automated ideation and reshaping design workflows to design education. Despite the increasing realism of synthetic imagery, several research gaps remain including alignment, plausibility, explainability, and control. This study focuses on alignment with human perceptions, specifically examining how synthetic architectural façade imagery aligns with geographical, objective, and affective text descriptors. We propose a pipeline that applies a Latent Diffusion Model to generate façade images and then evaluate this alignment through both AI-based and human-based evaluations. The results reveal that while images generated with geographical prompts are notably aligned, they also show clear biases. The results also reveal that images synthesised from objective descriptors (e.g., angular/curvy) are more aligned with human perceptions than affective descriptors (e.g., utopian/dystopian). These initial results highlight the opportunities and limits of current generative AI models, hinting at data biases and the potential lack of embodied understanding to grasp the complexity in experiencing architecture. Limitations of the study remain. Future work can expand on exploring cultural biases and semantic overlaps, and in testing more advanced embodied AI models and methods.

The study of interpretation is of major importance for our understanding of design cognition. When interacting with design representations, designers often rely on metaphorical descriptions as interpretive devices, which aid in coping with the task at hand. Consequently, such descriptions can enlighten us regarding the designer's perspective of the situation, and their analysis can deepen our knowledge of design cognition. We observe designers as they metaphorically interpret design representations during a simple task of spatial configuration, and introduce an approach for modeling this practice, as a means for getting insights into the designer's mental world. In this, we draw on traditional practices of protocol analysis, as well as on state-of-the-art theoretical frameworks for situated design and discourse analysis. Our integrated approach demonstrates how important relations between external and internal reality in design activity can be mapped and visualized. This sheds some light on the cognitive process of interpretation in design. The proposed method can both serve as a basis for detailed analyses of design cognition and for the enhancement of current models for situated design agents.

Generative AI is increasingly applied in architectural research, from automated ideation and reshaping design workflows to design education. Despite the increasing realism of synthetic imagery, several research gaps remain including alignment, plausibility, explainability, and control. This study focuses on alignment with human perceptions, specifically examining how synthetic architectural façade imagery aligns with geographical, objective, and affective text descriptors. We propose a pipeline that applies a Latent-Diffusion-Model to generate façade images and then evaluate this alignment through both AI-based and human-based evaluations. The results reveal that while images generated with geolocation prompts are notably aligned, they also show regional biases. The results also reveal that images synthesized from objective descriptors (e.g., angular/curvy) are more aligned with human perceptions than affective descriptors (e.g., utopian/dystopian). These initial results highlight the opportunities and limits of current Generative AI models, hinting at data biases and the potential lack of embodiment/experience/memory of these models to grasp the complexity in experiencing architecture.