Academic journal article Journal of Electronic Commerce Research

Improving Agreements in Multi-Issue Negotiation

Academic journal article Journal of Electronic Commerce Research

Improving Agreements in Multi-Issue Negotiation

Article excerpt


In bilateral negotiations with multiple issues, it is difficult to reach Pareto optimal outcome when the agents' preferences and relative importance of the issues are not known to each other. Self-interested agents often end up in inefficient agreements. Existing negotiation frameworks suggest solutions using trusted mediator, where agents will reveal their true preferences to the mediator. But in real life situations, a mediated solution is not always preferable, e.g., agents may not trust a third party to reveal their preferences. Without preference revelation, agents cannot improve the efficiency of the outcome. This paper presents an extended protocol for bilateral multi-issue negotiation. We show that with this protocol self-interested agents are able to explore and reach win-win agreements without revealing its complete preference.

Keywords: Software agents, multi-issue negotiation, efficiency.

(ProQuest Information and Learning: ... denotes formula omitted.)

1. Introduction

Both in human society and multi-agent society, negotiation is the main and most well known approach to resolve a conflict. With the massive growth of E-commerce, the research in automated negotiation is becoming increasingly important [Jennings et al. 2001, Kraus et al. 1995]. In automated negotiation, agents negotiate the contracts on behalf of the real life negotiating parties they represent. An agent negotiates over resources or services with other agents or humans. Many of these negotiations involve bargaining over multiple issues. The issues may be correlated or not. Here we have considered the issues are not related, that is, the utility of one issue does not depend on any other issue. We consider a negotiation scenario where two agents are negotiating over a multi-dimensional resource. Here, each dimension represents an issue. The preference of two agents may vary, i.e., one agent may prefer one issue over other while the other agent may prefer the opposite. Even in case of same order of preferences, the weights may vary. Here, the interaction between agents is not repetitive. No agent has any information about the other agent's preference structure. This is a reasonable assumption as with a rapid growth of electronic commerce, size of the market becomes so large that for some domain of e-negotiation it is not possible to know an individual and its preferences.

In multi-issue negotiation, agents with different preferences can cooperate each other to reach agreement that is beneficial for both the agents. We call a negotiation outcome efficient if it is Pareto optimal. A Pareto-optimal solution is one, where the utility of one agent cannot be improved without decreasing other agent's utility. Consider a situation where two children is dividing two cakes kept on a table, one made of chocolate and the other made of strawberry. Both of them like both cakes but the first child likes chocolate more than strawberry and the other prefers the strawberry to the chocolate. If they cannot reach an agreement they get nothing. Then an optimal solution is that both the children get the cake they prefer most. But if the agents do not know each other's preferences, then both will think, for each of the cakes, if the other guy is not offered at least half of the cake, he will not accept the offer. So, self-interested agents often fail to exploit the opportunity and end up in an inefficient "equal split" agreement. Note that, none of them has any incentive to sacrifice any portion of the cake unless it receives some which it prefers at least as much as what it relinquishes.

In the existing literature, there are two types of bargaining framework available for multi-issue negotiation. One is simultaneous framework, where, all the issues are negotiated simultaneously [Raiffa 1982] and the other is issue-by-issue negotiation, where, agents negotiate the issues one at a time [Bac and Raff 1996, Inderst 2000]. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.