Linking Socio-Spatial Practices to Density in Traditional Settings Using a Neural Networks Approach

The links between population, building density, and socio-cultural practises are rooted in the ways in which people interact with their built environment and with each other. Research to uncover these interactions and transformations has used various methods, including historical research, community participation, phenomenological interpretations, and spatial analysis. However, recent technological breakthroughs in artificial intelligence (AI) have had a considerable impact on the visualisation of density and its impact on associated socio-spatial practices; they have made it easier to collect and analyse field data and to produce visual representations of that data. Advancements in the study of urban environments using artificial intelligence (AI)—neural networks in particular—have the potential to have a substantial impact on our understanding of tradition as reflected in sociocultural practise and the physical/urban environment. This study presents a neural network-based approach to socio-spatial practises and density in Mumbai to understand how spatial behaviours along main pedestrian environments have transformed with the increase of urban density in selected well-established, traditional settings. This approach using neural networks identifies: 1) patterns and linkages between urban density and the occurrence of socio-cultural activities; 2) cultural hotspots based on the density and intensity of socio-cultural practises; and 3) the variation in urban density affects the character of the areas.

The research strategy includes four key steps: data collection, pre-processing, neural network training, and analysis. Collecting field data on population density, urban fabric, and sociocultural behaviours is the initial step. This data is structured to establish linkages between density, urban fabric, and socio-cultural behaviours in preparation for preprocessing. The second step is representing urban fabric and density as images and spatial data, as well as sociocultural behaviours as labelled or quantitative data. In the third stage, neural networks are trained using the Image Segmentation Algorithm, which is a type of convolutional neural network (CNN) that captures spatial relationships and hierarchical representations of visual characteristics. Using the trained CNN (pattern recognition), the fourth phase will analyse the variation in patterns and correlations between socio-spatial practise and urban density to determine the factors influencing the nature and frequency of cultural activities.

This research yields three outcomes: 1) the impact of urban density on socio-spatial practices; and 2) the identification of potential threats posed by rapid densification to socio-cultural practises in traditional environments. 3) focused preservation measures to safeguard sociocultural practises in light of urban density. Overall, the empirical findings can inform strategies that encourage spatial diversity and safeguard traditional environments. While this method yields accurate results, it has several drawbacks, such as the possibility of information bias and the inability to generalise the findings to a variety of urban situations in a conventional setting. Nonetheless, the method provides a starting point for employing AI to examine the relationships between population, building density, and spatial behaviours in conventional contexts.

Track: The Dynamism of Socio-Spatial Practice and the Making of Built Environments.
Session A10: Socio-Spatial Transformation.