A Coasean Democracy

Suppose residents of a simulated street disagree about evening deliveries. A shop needs access, some residents need quiet, and a delivery service wants a predictable schedule. Give each resident an advocate agent to explore possible agreements on their behalf.

The attraction is practical. An agent could find affected parties, compare acceptable schedules, and keep track of commitments. These activities involve transaction costs: the time and effort of finding, negotiating, monitoring, and enforcing an agreement.

Coase’s work provides a benchmark for thinking about those costs. In the idealised case of clearly defined rights and costless bargaining, parties can reach an efficient allocation; who receives the gains still depends on their starting rights. Real bargaining has costs, so the institutional arrangement matters.

For this proposal, I would protect some decisions from continuous bargaining. Basic rights would be fixed by a constitutional framework. Each resident would have equal civic standing in democratic decisions. Negotiation over delivery times would happen inside that framework, with a way to challenge an agreement and a body responsible for enforcement.

Begin with an imperfect representative

The advocate agent would have an incomplete account of its resident’s preferences. It might treat a preference for quiet as absolute when the resident would accept occasional deliveries, or trade away something the resident considers essential. A fast agreement can still be a bad representation of the person who must live with it.

I would compare five arrangements in the same proposed street simulation: periodic voting, elected representation, direct agent bargaining, liquid delegation, and bargaining within a protected set of rights. Liquid delegation lets a person vote directly or pass their decision to a representative, with the ability to change that delegation.

The runs would vary negotiating costs, access to resources, and how accurately agents understand residents’ preferences. They would allow coordinated strategies. One participant could create multiple apparent identities—a Sybil attack—or a coalition could withhold agreement to seek a disproportionate benefit.

Judge the agreement from the resident’s side

I would measure whether residents accept the final arrangements when shown their consequences, which groups repeatedly lose out, how much effort bargaining requires, and whether decision-making concentrates among a few representatives. Stable rules and a usable appeal process belong alongside speed in that assessment.

The proposal draws on arguments about lower coordination costs in agent systems, including those by Séb Krier and Herbie Bradley. Their implications pull in different directions: cheaper negotiation could widen participation, while advantages in data, resources, and organisation could concentrate power. The comparison above is intended to make those competing possibilities observable.

A successful negotiation would need to remain acceptable to the people represented after they understand the bargain. That is the standard I would start with.

References

Ronald H. Coase, “The Problem of Social Cost” (1960) and Nobel Prize lecture (1991); Séb Krier, “Coasean Bargaining at Scale” (2025); Alan Chan et al., including Herbie Bradley, “Visibility into AI Agents” (2024); Herbie Bradley, “Coasean economics of agent swarms” (2026).