Learning from the Metaheuristics: Protocols for Automated Negotiations

• Published in: Group decision and negotiation Vol. 24; no. 2; pp. 299 - 332
• Main Authors: Lang, Fabian, Fink, Andreas
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.03.2015; Springer Nature B.V
• Subjects: Algorithms; Automation; Bargaining; Biological and Physical Anthropology; Business and Management; Commodities; Contract negotiations; Decision analysis; Decision making; Heuristic; Heuristics; Imperfect information; Learning; Management decisions; Negotiations; Protocol; Software; Strategic planning; Studies; Utility functions; Germany; Automated negotiation; Negotiation protocol; Complex contracts; Mechanism design
• ISSN: 0926-2644; 1572-9907
• DOI: 10.1007/s10726-014-9390-x

Nowadays, enterprises are more and more interconnected such that operational planning has to consider the different interests of the involved organizations. This may be a challenging and complex task as it is subject to strategic interactions and incomplete information. Automated negotiation by software agents is a powerful tool which can handle these issues and facilitate intercompany planning. Nevertheless, sophisticated negotiation protocols that govern the rules of the negotiation are needed. In this study, we present and evaluate two configurable protocols for multi-issue negotiations, which are inspired by general heuristic optimization algorithms for centralized problems, so-called metaheuristics. The protocols consist of several policy building blocks; these are evaluated with regard to their impact on the negotiation outcome. The evaluation shows that both protocols can efficiently achieve beneficial solutions—even for complex, nonlinear contract spaces—given the parameterization and the configuration of building blocks are chosen appropriately. Furthermore, we elaborate on requirements for appropriate protocol design and find that both protocols adequately comply with the requirements.