Regional Policy and Solo Founders Can Now Hear Each Other: A Mediating Agent Lifts the Government Objective 42.4% Above the Eight-Baseline Mean
Synopsis
Chinese regions use separate support policies to compete for one-person companies, solo founders who work through digital platforms and AI agents and can compare locations, and the two sides used to act on their own separate estimates. PolicyMediator turns that into a cognition-constrained leader-multi-follower Stackelberg framework: a mediator agent explains verified policies to the companies to build awareness and trust, then reports their needs and unresolved concerns to governments for policy learning. Across six-region simulations in China with five OPC types, five scenarios and three simulator backbones, it ranks first in all 15 combinations, with the aggregate government objective 42.4% above the eight-baseline mean.
Interpretation
Policy learning now carries what the companies understand about a policy back to the policy side, instead of leaving each side to work from its own estimate. Earlier policy-learning methods optimized resources under fixed perceptions, and consensus-formation methods pooled views without feeding the resulting cognitive change back into policy learning, so location choices and regional policy adjustments mismatched. Measured on the aggregate government objective, which combines economic performance, employment benefit and policy-access equality, PolicyMediator improves on the mean of eight baselines by 42.4%, and raises average OPC utility by 40.9%.
The mediator carries messages both ways and revises its own strategies: it explains verified policies to the companies to build awareness and trust, reports their needs and unresolved concerns to governments, and keeps, revises or prunes strategies using verification, allocation, capacity and objective trajectories rather than mediation success alone. The arrangement is a cognition-constrained leader-multi-follower Stackelberg framework, with governments as leaders and one-person companies as followers, and the mediator relaying between them. Testing ran on six-region simulations in China, where five OPC types, five scenarios and three OPC-simulator backbones form 15 combinations, and PolicyMediator ranked first in every one.
Perspective
The bilateral mediation was tested entirely in six-region simulations in China, covering five OPC types, five scenarios and three OPC-simulator backbones, with scores on the aggregate government objective and average OPC utility. The demonstrated scope is simulation, written for researchers working on agent-mediated mediation, multi-agent negotiation and regional policy design.
Misperception is the premise the design addresses: one-person companies may reject otherwise suitable locations because they understand a policy imperfectly, and the summary reports only simulation results on whether mediation resolves that. In the tests the companies are played by three OPC-simulator backbones rather than real entrepreneurs, which limits how far the result carries. This report reads the summary, so method details and the full experimental setup were not checked line by line.