Research On A Novel Bargaining Protocol For Open E-Marketplaces

Alternating-offer bargaining protocol (AOP) is the most predominant way for solving bilateral bargaining problem in daily life. Almost all past theoretical and practical researches on bargaining problem use AOP as the default interaction rule. However, AOP is not so perfect because it has at least two shortcomings that have not been considered much enough by predecessors of the field. The first one is "Alternating-offer"may cause some kind of potential benefit lost of the bargainers, and the second is that some speculative bargainer could abuse the aspect of "Alternating-offer"to get a better bargaining position or impair his opponent's benefits. Since these flaws have great impacts on aggravating bargainers'suspicions and mistrust about each other, they are likely to make the negotiation process much slower, i.e. consuming more time and causing a lower efficiency. Being different from human being, software agent has its intelligence embodied entirely by logic and reasoning. Except for the powerful computing capability, agent is endowed with some basic intelligence to fulfill special tasks, which are supposed to be done by human being. But we do not need the agent to be a full human-like entity. In fact, with our current knowledge and technology level, those extremely complicated characteristics of humankind, such as intuition, inspiration, happiness, sadness, hesitation etc. are still impossible to be realized for software agent. In this sense, it is safe to say that agent just is a kind of tool with a given special capability. Based on this view, we pointed out two important aspects by which agent bargaining differs from human bargaining: firstly, agent is almost fully rational. Its behavior is completely determined by logic, reasoning, rules, beliefs, desires and intentions. As we know, people are easy to be distracted by their mood when making some decision or action, while agent doesn't have this problem, because it can always deal with uncertainty or fuzzy information through certain probability or degree of membership so that it can make its decision in a quick and explicit manner. Secondly, the bargaining between agents is structuralized and formalized. It is not as free as the bargaining occurs between human beings. During the bargaining process, agent must comply with the preset interaction rules. Besides, agents communicate through formalized language, and the messages transmitted between them must be well structuralized and formalized too. According to the shortcomings of AOP and the particularities of the agent bargaining, we believe agents should not be restricted to AOP for their requirement of bargaining, therefore, a novel bargaining protocol for agent world, which is called Sealed-Offer Synchronous Bargaining Protocol (SOSBP), is presented in the paper. SOSBP overcomes the shortcomings of AOP through a middleman's intervention and a set of concession-related rules. Under SOSBP, two bargainers do not need to alternately submit their offers or counter-offers any more, for all their offers are sent only to the middleman and kept invisible from each other. The middleman's duty is to lead the bargaining process moving on. As the process is a multi-stage negotiation, in any single stage, the middleman will collect bargainers'offers and then check out if they match each other. If matched, the middleman will compute the final price according to related rules of SOSBP, and then terminate the bargaining session; if mismatched, it will inform both bargainers with the outcome of current stage and wait for their next moves. The process will continue until an agreement is reached or any bargainer decides to quit. We presented a comprehensive description for the main principle of SOSBP, and explained why it can prevent or avoid speculation or fraudulence during the bargaining process. Moreover, we also described some obvious differences between SOSBP and AOP from several aspects, such as bargaining procedure, deal price and changes on agreement zone. To analyze agents'strategies under SOSBP, we built a formalized bargaining model S-AB. It is composed of bargaining information, beliefs, action space and expected utility etc. Although we can treat the S-AB model as a particular dynamic game with incomplete information, the bargaining strategy under SOSBP can not be analyzed by the means of dynamic game study, because some common-knowledge assumptions, which are required when analyzing incomplete information game, are removed from S-AB model. Therefore, the bargainers do not need to consider the cross-influence between them caused by the common-knowledge assumptions, and they will make decisions only according to their private bargaining information and the beliefs about their opponent's private info. In this way, we treated the S-AB model just as an individual decision-making model, and thenanalyzed the bargainer's strategy respectively from buyer and seller's viewpoint. E-Market is an important branch of E-commerce, and it is the main area which SOSBP can be applied to. As a novel bargaining service provided in E-Market, SOSBP can help the involved traders ravel out their disagreement on price more quickly and effectively. As MAGNET is a market-based platform for multi-agent negotiation, which provides support for a variety of negotiation type, it is reasonable to incorporate SOSBP with the MAGENT framework. Basing on the related literatures of MAGNET, we finished some fundamental work for the specifications of main actions and messages required by SOSBP. We also worked out a draft of necessary components of the bargaining agents, and explained the correlativity among them. In order to test the feasibility and effectiveness of SOSBP, an experimental study is carried out for comparing the performance of SOSBP and AOP. The issues about experimentation settings, the bargaining data generation and the evaluation measures are respectively discussed in the paper. By analyzing and comparing the experimentation outcome, we draw a positive conclusion just as our initial expectation, namely SOSBP has better performance than AOP at least in three aspects: time saving, deal rate and fairness. Besides the traditional protocol, SOSBP could be an alternative for the bilateral price negotiation between two trading agents and make the process going faster and more smoothly.