Façade is the face of a building and the connection between inner and outer spaces. It gives a building its identity and image and influences how people respond to it physiologically and behaviourally. Architects spend significant effort in designing building facades, often producing various design iterations and options for discussions with clients. Design decisions on façade are often varied and subjective. The paper examines the design strategies used in high-density public housing facades in Singapore and propose an approach to understand their effects on visual preference through quantitative studies. The study involves two approaches in the extraction of visual indices from facades – one using rule-based 3D parametric modelling for synthesis, while the other using deep convolutional neural networks for prediction. To examine their visceral responses, subjects are asked to view different façade designs in a controlled environment while wearing biosensors. The façade designs are presented either as non-photorealistic images rendered from different perspectival views or as photorealistic imaged generated with stable diffusion from different front elevational views. The findings are framed as a methodological contribution, through a combination of existing parametric modelling and recent deep learning approaches, to the area of cognitive architecture and visual preference.