Arbitration Protocols for Competing Software Agents

The exchange of goods and services between bargaining software agents requires new forms of brokering mechanisms which achieve consensus between conflicting pal-ties. Such mechanisms have to be designed in a way that they give rational self-interested agents no incentives for insincere behavior. We introduce an arbiter as third party that resolves conflicting bargaining situations between the agents. To achieve non-manipulative agent behavior, we investigate three arbitration protocols that avoid different forms of manipulations and show how each trades net efficiency for robustness against manipulations. We describe the applicability of the protocols in bilateral bargaining situations and, analyze their robustness against manipulations analytically and by simulations. We compare the protocols with Nash's arbitration(1) and the Groves-Clarke tax(2) and characterize situations in which our protocols are superior.