Zero-Carbon Futures is a speculative, AI-aided research project that supports participatory and data-informed reimagining of urban environments in response to climate change and evolving mobility patterns. The project critiques prevailing demolition practices, which often overlook the embodied carbon within existing structures and fail to consider the environmental and social value of adaptive reuse. Instead, it promotes a low-carbon, inclusive urban future through community engagement and computational design tools.
At its core, the project introduces a digital platform that integrates generative AI and large language model (LLM)-based agents to facilitate real-time, site-specific design feedback. Users interact with digitized 3D models of existing urban spaces—such as neighborhoods or infrastructural sites—generating speculative design proposals via natural language prompts. This design-by-dialogue approach lowers the barrier to entry for non-expert users while still allowing complex, site-specific scenarios to be explored in depth. Community members can re-envision their local neighborhoods using simple prompts.
The platform uses a multi-agent AI system, where different LLM-based agents simulate the perspectives of a range of stakeholders—urban planners, transportation experts, environmental scientists, and local residents. These agents evaluate proposed design interventions across qualitative spatial metrics such as walkability, greenness, safety, beauty, and monotony. Each agent uses an evaluative framework based on urban design literature and publicly available datasets to assess proposals and provide feedback.
On the quantitative side, machine learning algorithms analyze and classify existing and proposed structures using carbon-related metrics. This includes estimating embodied carbon, operational carbon, and potential emissions savings from reuse versus demolition. Computer vision tools process 3D models and site images to detect material types and construction typologies, while clustering algorithms segment this information based on carbon intensity.
These layers of analysis are visualized in real time, helping users understand the implications of different design choices on carbon outcomes and other sustainability indicators. The platform also supports scenario comparison, allowing users to toggle between different futures and evaluate trade-offs across environmental, social, and aesthetic dimensions.
The methodological foundation draws from participatory planning, computational design, and human-AI interaction research. The project applies this methodology to a specific case study: the Quincy Adams Station Parking Garage in Quincy, Massachusetts. As a prototypical car-centric infrastructure with a high carbon footprint, the garage serves as a critical site for testing the platform’s potential. The case study imagines a future in which private car ownership declines due to the rise of autonomous vehicles, shared mobility systems, and increased investment in transit-oriented development. In this context, the garage—currently a large, carbon-intensive monolith—becomes a canvas for community-led design exploration.
Using the platform, community members and stakeholders engage with the site’s 3D model to propose adaptive reuse strategies—ranging from mixed-use housing to urban green and community spaces. AI agents provide site-specific feedback, while machine learning models estimate carbon impacts. This AI-augmented co-design process empowers communities to make informed, creative, and sustainable decisions about their built environment.
Zero-Carbon Futures bridges computational tools, urban policy, and citizen participation, offering a new model for AI-integrated, community-led design. It demonstrates how speculative design, supported by technical workflows and inclusive engagement, can help transform high-carbon infrastructures into resilient, future-ready spaces.
George Guida
Daniel Escobar
Tatjana Crossley
Giovanna Pillaca
Carlos Navarro
Jean Santos
Andrew Witt