CoRenew uses LLM agents to simulate multifamily housing redevelopment policies

A new open-source platform called CoRenew, created by Yudi Zhang and colleagues, applies large language model (LLM) agents to tackle the classic collective action problem in multifamily residential redevelopment. Traditional simulation models rely on predefined behavioral rules, which fail to capture how stakeholders adapt their decisions as negotiations evolve. CoRenew replaces those static rules with LLM-powered agents that dynamically adjust strategies based on changing contexts and other agents' reactions. The platform ingests open-source geographic and demographic data to generate synthetic residents and simulate alternative policy combinations, supporting both numerical and semantic policy inputs. It includes built-in visualization and export tools, making it practical for policy analysts.

CoRenew's behavioral realism was validated against survey responses from 324 real residents and a nine-month observed negotiation process from an actual redevelopment project. The modular architecture allows adaptation to different institutional and cultural contexts. By enabling ex-ante evaluation of how policies influence stakeholder behavior, CoRenew offers a significant advance over static models. It can help policymakers test incentives, timing, and targeting strategies to overcome barriers to redevelopment, potentially accelerating housing renewal projects and improving community outcomes.