Multiple Issue Negotiation Based on Standardized Documents

An evolving approach to conduct business transactions over B2B electronic marketplaces is to temporarily connect the ERP systems of the involved business partners. For this type of B2B integration, standardized processes, interfaces and business documents as provided by, e.g., RosettaNet or the NES Consortium constitute an indispensable requirement. However, the respective standards have significant deficits regarding the negotiation of contracts such as purchase orders. The work at hand closes that gap by providing a novel XML communication framework for argumentation-based negotiation which incorporates standardized business documents and allows for interaction between human as well as software agents. Concrete application scenarios are outlined. 1 Electronic Negotiation and Standardization Efforts There has been a lot of work done in developing middleware for integrating and automating business processes across company boundaries. Recently, XML based technology approaches gained increasing importance in this area. Various consortia and organizations (e.g. RosettaNet1, NES Consortium2) offer XML based business document definitions and process specifications covering many aspects of business within companies of all sizes. However, these standards have significant deficits regarding the negotiation of agreements and are often little more than an electronic catalogue from which a customer can select and order goods. This may be appropriate to many scenarios, but other ones certainly require more sophisticated mechanisms to computationally negotiate, e.g., price, quality and delivery terms of a purchase order (Turowski 2002). In contrary, multi-agent negotiation systems based on communication languages such as FIPA’s ACL (FIPA 2003) or KQML 1 http://www.rosettanet.org/ 2 http://www.nesubl.eu/ Christoph Pflügler 1204 (Finin et al. 1994) are fairly complex and, as Beer et al. (1998) state, too academic for many applications. The work at hand strives to close that gap by providing a communication framework for bilateral negotiation which is based on the latest findings in electronic negotiation and argumentation theory, and allows for a seamless integration into existing standards such as RosettaNet’s PIPs or the NES profiles through its ability to negotiate on the basis of pre-existing, standardized XML document definitions. The paper is structured as follows: Section 2 gives a brief overview of the field of electronic negotiation and thereby lays the theoretical foundation for the framework presented. Related work is summarized in section 3. Section 4 elaborates on the requirements for and describes the core components of the proposed communication framework for argumentation-based negotiation: The structure of a so-called negotiation extension for pre-existing business document definitions and a bilateral negotiation process defining required tasks, the control flow of each participant, as well as the resulting message exchanges. An exemplary incorporation of the negotiation extension into an XML document definition is depicted. Further, the framework is evaluated against requirements on negotiation mechanisms. The contribution of the communication framework presented by the work at hand is outlined in section 5, which also summarizes ongoing and future work. 2 Theoretical Background Negotiation theory distinguishes two types of negotiation (Hung et al. 2004; Walton and McKersie 1965): competitive negotiation (also referred to as zero sum or distributive negotiation) and collaborative negotiation (integrative negotiation). The first one can be characterized by the fact that every negotiating party exclusively tries to maximize its individual utility. In contrary, collaborative negotiation aims at finding and adopting the solution that provides the greatest joint utility taken the parties collectively. These two types represent two extremes of the spectrum of possible negotiation types; most real-world negotiation scenarios contain competitive as well as collaborative elements (Walton and McKersie 1965). Further, electronic negotiation is defined as a joint decision-making process of two or more parties within an electronic market (Smith and Davis 1983). It can be seen as an optional activity within the agreement phase of an electronic market transaction (Rebstock et al. 2003). Electronic negotiation essentially deals with three topics: negotiation protocols, negotiation objects and the agent’s decision making model. According to Jennings et al. (2001), a negotiation protocol is a set of rules that govern the interaction. This includes the permissible type of participants such as the negotiators and relevant third parties, negotiation states, events that trigger a change of the negotiation state, permissible actions of each participant in a particular negotiation state, as well as resulting message exchanges. Regarding the design of a negotiation protocol, there exist two fundamental alternatives: offer-counter-offer protocols (also called MKWI 2010 – E-Commerce und E-Business 1205 send/resend protocols) on the one hand and mediated protocols on the other. During a negotiation performed in the offer-counter-offer fashion, one party offers a proposal to the other party, which in turn accepts the proposal or sends a counterproposal. This procedure is repeated until an agreement is reached or a party aborts the negotiation. In the context of the work at hand, the term mediated protocol does not refer to mediators used to ensure quality of service properties such as non-repudiation or mediators that serve as message queues to store and forward negotiation messages. Mediated protocols as the one proposed by Klein et al. (2003) can be leveraged for fully automated negotiation in a multi-agent environment. They employ a central mediator that initially generates a random contract. Each negotiating party then votes to accept or reject the proposal. If the proposal is accepted by both parties, the mediator mutates that proposal (by randomly flipping a value of an issue) and repeats the procedure. In case one or both parties vote to reject, a mutation of the last mutually accepted proposal is suggested instead. This procedure is repeated for a fixed number of steps. Negotiations can further be categorized by negotiation objects which constitute the range of issues an agreement must be reached for (Jennings et al. 2001). Single issue negotiations, such as negotiating a price only and multiple issue negotiations, e.g., negotiating price, quality and delivery terms are distinguished. If automated agents are involved in the negotiation process, their decision making model is utilized for further classification (Jennings et al. 2001). On the one hand, game theoretic models aim at finding the best solution for a negotiating party. However, they are subject to fundamental limitations, most notably partial knowledge, inconsistent beliefs and bounded rationality (Först et al. 2009). On the other hand, bargaining agents incorporating heuristic approaches try to find a good instead of the optimal solution and overcome the limitations of game theoretic approaches by searching the negotiation space in a non-exhaustive manner (Jennings et al. 2001). Moreover, argumentation-based approaches have been proposed as an alternative to classical mechanism design (Först et al. 2009). They extend simple proposalbased heuristic approaches by the ability of an agent to justify and persuade the counter-party of its negotiation stance by the use of arguments.