Detecting similar negotiation strategies

Abstract

Automated negotiation is a key form of interaction in complex systems composed of autonomous agents. Negotiation is a process of making offers and counteroffers, with the aim of finding an acceptable agreement [1]. The agents (negotiators) decide for themselves what actions they should perform, at what time, and under what terms and conditions [1, 2]. The outcome of the negotiation depends on several parameters such as the agents' strategies and the knowledge which one agent has about the opponents [2-5]. In recent years, the problem of modeling and predicting negotiator behavior has become increasingly important because this can be used to improve negotiation outcome and increase satisfaction of results [2-6]. In this chapter we consider the problem of defining strategies' similarity or distance between strategies. We start with the idea that similarity between negotiators should somehow reflect the amount of work that has to be done to convert one negotiation session to another. We formalize this notion as Levenshtein or edit distance [8, 9] between negotiations. We apply dynamic programming for computing the edit distances and show the resulting algorithm is efficient in practice. Indetail, the chapter is organized as follows. In Sect. 22.2 we present the problem in negotiations. The definition of similarity between negotiation strategies is given in Sect. 22.3. In Sect. 22.4 we review the negotiation protocol used in our experimentation. We use some negotiation strategies in our simulation discussed in Sect. 22.5. In Sect. 22.6 we present some results of computing similarity measures. Section 22.7 contains conclusions and remarks about future directions. © 2008 Springer Science+Business Media, LLC.