How can an agent learn to negotiate?

Abstract

Negotiation has been extensively discussed in game-theoretic, economic, and management science literatures for decades. Recent growing interest in autonomous interacting software agents and their potential application in areas such as electronic commerce has given increased importance to automated negotiation. Evidence both from theoretical analysis and from observations of human interactions suggests that if decision makers can somehow take i n to consideration what other agents are thinking and furthermore learn during their interactions how other agents behave, their payo might increase. In this paper, we propose a sequential decision making model of negotiation, called Bazaar. It provides an adaptive, multi-issue negotiation model capable of exhibiting a rich set of negotiation behaviors. Within the proposed negotiation framework, we model learning as a B a yesian belief update process. In this paper, we present both theoretical analysis and initial experimental results showing that learning is bene cial in the sequential negotiation model.

Negotiation decision functions for autonomous agents

“…This is because the initial o ers of the server are now i m m e diately within the acceptance ranges of the client (con rmed by t h e n umber of messages exchanged before a deal is reached ( gure 10.D). Thus the client gains relatively more utility than a server since the initial o ers of both agents are low and deals are made at low v alues 16 .…”

A Service-Oriented Negotiation Model between Autonomous Agents

“…Contrary to our model, and as we discussed in Section 2, this is an inappropriate assumption from the point of view of real applications. As has been argued elsewhere [10], these and other assumptions limit the applicability of game theory models to solve real problems.…”

“…Hence, our study, and those forthcoming, are much more in the experimental line of [4]. Although we do not concentrate on learning, some similarities can be found with the formalism by Zeng and Sycara [10]. We have not concentrated however on the description of negotiation protocols that has been an important focus of attention for the community of distributed artificial intelligence (see [7] for extensive references).…”

Designing and implementing a multi-agent architecture for business process management

“…This protocol does not strictly involve negotiation (according to our definition) because bidding agents have only one shot at the available sub-problem. Richer models that address this criticism are the recursive negotiation model [14] and the sequential decisionmaking model [25]. In the recursive negotiation model [14], agents exchange information and plans, and produce critiques of other's proposed plans in an attempt to find a mutually acceptable solution to the distribution of action and control in solving the problem.…”