A Metanegotiation Support System: Supporting Protocol Search between Rival Agents

Abstract

Designing successful negotiation protocols for particular negotiation scenarios is a very challenging task. Negotiations not only have to cope with strategic behavior by the negotiators but also with restricted and false centrally available information. Recent research covers this design problem and proposes a variety of potential procedures. However, there is no single protocol that is the answer to everything. Furthermore, the protocol choice can have a significant influence on the outcome of the negotiation. Since there is no benevolent mechanism designer in real-worlds applications, the negotiators have to find an agreement on the rules of encounter on their own. In consequence of self-interest, the protocol search constitutes a negotiation problem in itself. This paper introduces and discusses the problem of metanegotiation. A metanegotiation represents a higher-level negotiation for a protocol which is used for an underlying lower-level negotiation. To support the parties in finding adequate protocols, we present an architecture for a metanegotiation support system (MNSS). The MNSS comprises different tools such as logs, repositories, or analytical tools that can also serve as a stand-alone decision support system for the agents. The framework features ex ante and live support, protocol selection and generation, and different learning approaches. Moreover, we develop preliminary classifications for negotiation scenarios as well as protocol rule categories and show how related tools from literature can be integrated in the system. Finally, we give an overview on how we intend to approach the presented problem in future work.