22: Mapping Aesthetics In Competitions: Applying A Computational Aesthetics Framework For Historical Comparison of Architectural Competitions

This study applies a computational aesthetics framework to analyze and compare architectural competition submissions for the Brazilian pavilion at five World Expositions (1970–2025). Unlike previous studies focusing on winning entries, this research includes a broader range of submitted proposals, enabling historical analysis of formal tendencies over five decades. A dataset of perspective images was compiled from academic sources and official archives. Six aesthetic metrics—compression complexity, brightness, fractal dimension, depth composition, aesthetic measure, and predicted hedonic response—were extracted using image processing, segmentation, and a trained neural network model. The results reveal evolving design trends, including increasing visual complexity, stabilization of brightness, and a return to spatial contextualization in recent renderings. Surprisingly, aesthetic efficiency and predicted pleasure remain loosely correlated, suggesting designers balance compositional order with emotional resonance. This research demonstrates how computational methods can enhance our understanding of historical aesthetic shifts in architectural representation and competition culture.