Predictive ‘Devices’ For Planners: A Critical Review From Morphogenetic Models To AI

Since the late 1970s, urban modelling has often drawn on morphogenetic principles, employing tools like cellular automata, multi-agent systems, and fractal geometry to reduce complex urban phenomena into simplified, computable forms. While such reductive models are valuable, they risk oversimplification - especially in fields like urban design, which behave as “non-trivial machines” where outcomes are not fully predictable from inputs. This unpredictability demands intelligence, adaptability, and the capacity to respond to unforeseen conditions during the design process. Despite their limitations, data-driven tools and mathematical models can enhance designers’ understanding, offer novel visualisations, and support engagement with high-resolution data. In this context, we propose a shift toward sustainable, data-network-based modelling approaches that address specific environmental challenges, such as urban heat islands. These issues require tailored, adaptive models that evolve with real-time input and continuous training. Here, Artificial Intelligence (AI) becomes crucial. Its capacity to process vast data streams - collected, for instance, via sensor networks - can help identify patterns in urban environments, assess the impact of recent interventions (e.g., urban reforestation or depaving), and support the development of more accurate and responsive models. Moreover, AI’s generative capabilities allow planners to explore multiple scenarios, including unconventional ones, enriching design processes without replacing human agency.