Negotiating agents

Catholijn M. Jonker*, Koen V. Hindriks, Pascal Wiggers, Joost Broekens

*Corresponding author for this work

Research output: Contribution to JournalArticleAcademicpeer-review

Abstract

Everyone is an experienced negotiator and everyone has an opinion about his or her negotiation skills. However, even professional negotiators can still improve their skills considerably. "Most people are ineffective negotiators ⋯. Fewer than 4 percent of managers reach win-win outcomes when put to the test .⋯ Even on issues for which people were in perfect agreement, they fail to realize it 50 percent of the time" (Thompson 2005). Although many forms of negotiation exist, in this article we focus on integrative bargaining (see Walton and McKersie 1965). Negotiation is a prime example of a task for which the human mind is but partially equipped, and for which artificial intelligence (AI) can provide assistance. Among others AI can provide search techniques, computational heuristics to tackle exponential problem spaces, strategic reasoning, argumentation, learning techniques, and affective computing to handle the complications that arise in negotiations. More difficult problems that are not as easily solved by artificial intelligence techniques, however, include obtaining the common knowledge necessary to understand negotiation domains and arbitrary human conversations that take place during negotiations. We aim for synergy between human and agent in such a way that the human weaknesses are covered by the strengths of the agent and the weaknesses of the agent are covered by the strengths of the human. This implies that tasks should be divided over humans and agents in a way that respects those capabilities. On the one hand, humans are better equipped to understand the context and the emotional fluctuations in human-human interaction, they are capable of finding new relations between concepts, and they have the necessary background knowledge to interpret the domain of negotiation with respect to their preferences. On the other hand, humans can be troubled by emotions, and have difficulty handling the complexity of negotiation spaces and keeping track of the interests of the negotiation opponent. For agents it is largely the other way around. The rest of this article is organized as follows. We first informally introduce human negotiation as a process of four phases that may be distinguished. We then proceed by discussing some state-of-theart negotiation support systems and automated negotiating agents. We then present the pocket negotiator project that is developing the first of a next generation of negotiation support systems. Thereafter we discuss various technical components based on different artificial intelligence techniques that are part of this agent, including support for analyzing a negotiation, taking the opponent into account, advice on how to negotiate strategically, for eliciting human preferences, and for handling emotions. The article concludes with an overview of open research questions.

Original languageEnglish
Pages (from-to)79-91
Number of pages13
JournalAI Magazine
Volume33
Issue number3
Publication statusPublished - 1 Sept 2012
Externally publishedYes

Fingerprint

Dive into the research topics of 'Negotiating agents'. Together they form a unique fingerprint.

Cite this