American cities face growing tension among stakeholders whose priorities often conflict. Residents seek affordable housing and livable neighborhoods, developers pursue economic viability, and policymakers must balance growth with equity and sustainability. These competing interests frequently surface through the NIMBY–YIMBY divide, where debates over housing and zoning reflect deeper disagreements about fairness, identity, and neighborhood change. Yet the mechanisms for dialogue between these groups remain limited and fragmented, leading to long planning cycles and a lack of shared understanding.
Agent City Hall (ACH) begins by asking how an artificial agent can think more like a person. Instead of merely imitating language, ACH reconstructs how people reason—how they form beliefs, weigh trade-offs, and adapt to new information. Each agent encodes a transparent chain of causal reasoning, allowing planners to see not only what people believe but why.
Public engagement, traditionally measured through surveys and town-hall feedback, often samples opinions without context and struggles to maintain up-to-date representations of the people it claims to reflect. Participation tends to be sporadic and biased toward those with time, access, or strong opinions, while many others remain silent or unseen. In practice, public hearings face empty chairs as much as vocal crowds, and the resulting data flatten social diversity into narrow statistical averages. When models extrapolate from such limited input, they risk out-of-distribution errors—amplifying dominant voices while overlooking minority reasoning patterns.
ACH rethinks this process by transforming engagement into an ongoing simulation of reasoning. Citizens are not static respondents but evolving cognitive models whose beliefs can be updated, contextualized, and recombined. By tracing how people reason rather than what they merely report, ACH captures the dynamics of public understanding and creates a participatory system that remains both inclusive and cognitively faithful over time.
By combining these cognitively grounded agents into a shared environment, ACH becomes an efficient consensus machine. It simulates how residents, developers, and policymakers debate and adjust their perspectives, revealing where values diverge and where common ground can emerge. Through this process, conflict turns into dialogue and planning into collective reasoning—making the future of urban governance more transparent, adaptive, and human-centered.