Humanizing Agent-Based Models (h - ABM)


Humanizing Agent-Based Models (h-ABM) emerges as a pioneering technique to simulate real-world behaviors, aiming to seamlessly bridge the computational realm with the intricacies of human dynamics through advanced algorithms and rich attributes. This methodology finds its profound resonance in future cities' design and predictive modeling, particularly under the aegis of the City Science research group.

Historical viewpoints on computational agents often depicted them as rudimentary, rule-bound entities, narrowly defined in scope and adaptability. Nevertheless, the contemporary paradigm, fortified by innovative AI breakthroughs, suggests a transformative potential: agents endowed with human-like attributes, creating a more vibrant and dynamic continuum in agent interactions. Humanizing ABM is not a mere academic venture; it has profound implications for the urban fabric, especially when seeking to address the profound challenges of tomorrow's cities.

In this framework, the City Science research group recognizes that novel strategies are indispensable for creating and evolving habitats and workspaces and the mobility systems weaving them together. AI and models like LLM are not just computational instruments in this endeavor; they emerge as vital catalysts. These tools augment our design and predictive capacities and offer an enriched, nuanced lens grounded in intricate cognitive modeling and real-world simulations. Through such AI-enhanced approaches, we foresee a near future where cityscapes and their inhabitants are better understood, predicted, and catered for. "Humanizing ABM" transitions from a conceptual framework to an actionable paradigm. As we push this frontier, technology and human essence come together, fostering urban environments that are sustainable and profoundly attuned to human nuances, ensuring cities that are more responsive, adaptive, and genuinely fit for the future.

h-ABM localization architecture

h-ABMs enhance ABMs with advanced cognition and LLM-driven actions for accurate human behaviour simulation.