Negotiation in exploration-based environment

摘要

When two parties need to split some reward between them, negotiation theory can predict what offers the parties will make and how the reward will be split. When a single party needs to evaluate several alternatives and choose the best among them, optimal-stopping-rule theories guide it as to how to perform the exploration, what to explore next and when to stop. We consider a model in which party A needs to choose one alternative, but has no information and no means of acquiring information on the value of each alternative. Party B, on the other hand, has no interest in what party A chooses, but can perform (costly) exploration to learn about the different alternatives. As both negotiation and exploration take time, the common deadline and discounting factor further tie the processes together. We study the combined model, providing a comprehensive game theoretic based analysis, enabling the extraction of the payments that need to be made between agents A and B, and the social welfare. Special emphasis is placed on studying the effect of interleaving negotiation and exploration, and when is this method preferred over separating the two. In addition to exploring the basic questions, we also consider the case in which one of the parties has some control over the parameters of the problem (e.g. the negotiation protocol), and show how it increases the utility of this party but decreases the overall welfare.