Menu
Informal and non-planned settlements are highly complex and typically develop through small-scale, bottom-up actions. Accordingly, their morphology results from environmental factors and reflects their specific social structure, economy, traditions, and cultural values (Habraken, 1998). Often regarded as chaotic and non-functional (Schaur, 1991), these settlements present a unique challenge to urban design practitioners and policymakers, since top-down policies are unable to follow their rapid bottom-up development. Adequate solutions to the substandard living conditions in non-planned settlements require a deep understanding of their initial formation and expansion processes and appropriate tools to evaluate potential policies or interventions (Patel et al., 2018).
To address this challenge, this research presents a machine-learning (ML) and computer vision approach that can be used as an urban design tool. The Arab town Jisr az-Zarqa in Israel was chosen as a case study. The lack of consistent planning policy in rural Arab settlements in Israel (Brawer, 1994) led to the unplanned development of Jisr az-Zarqa, which until 1988 had no official master plan (Israel Land Authority, 1992). A machine-learning model was trained to learn and reproduce the complex urban patterns of Jisr az-Zarqa, expressing the unique spatial qualities of this settlement.
This research argues that applying Artificial Intelligence (AI) to the design practice can be valuable due to its ability to learn urban patterns that result from dynamic and spontaneous self-organization processes. Therefore, AI tools can be meaningfully used by professionals, policymakers, and local communities to better visualize and reflect on the potential outcomes of different development scenarios. This research contributes a methodological framework that can generate development alternatives sensitive to the local practices, thereby bridging the gap between top-down and bottom-up practices, promoting the preservation of traditional knowledge and cultural sustainability.
