AGORA framework shows deliberation can neutralize transit participation bias

A new paper from MIT researchers Jung-Hoon Cho and Cathy Wu tackles a longstanding problem in transit network design: the people who show up to public hearings often don't represent the broader community, leading to biased outcomes. Their proposed framework, AGORA, treats participation bias not as an uncontrollable input but as a process-design problem. By simulating meetings with stakeholder agents, structured deliberation phases, and governance gates (filtering mechanisms that approve or reject proposals), the researchers could isolate how composition affects final transit plans.

Testing on two standard benchmark networks (Mandl and Mumford0) yielded three key findings. First, while aggregate outcomes varied little across compositions, skewed participant groups increased tail risk and fairness disparity — representative sampling consistently outperformed biased groups. Second, without any deliberation, composition had zero impact on outcomes, proving that deliberation itself is the mechanism through which who attends affects results. Third, governance gates effectively compressed cross-profile variance on the Mandl network without shifting average outcomes, but on the larger Mumford0 network, the low acceptance rate showed that threshold values require careful calibration per instance. The work reframes participation bias as a solvable process problem: well-designed deliberation and governance can substantially reduce how much public transit outcomes depend on who is in the room.