本研究應用多屬性決策概念於採購協商工作，並建構一個以網頁為基礎 (web-based) 的協商支援系統（negotiation support system, NSS），且考慮適當的協商者偏好、協商戰略與策略提出一協商流程，流程結束於共識的達成或是協商破裂。此協商支援系統中介於企業之間，例如買方與其供應商，以協助雙方在複雜決策情境下，遂行有效率及有效能之耐久財採購決策。 當買方在使用群體決策支援系統（group decision support system, GDSS）進行供應商遴選後，本中介協商支援系統支援買方與最高排序之賣方間協商的進行。過程中買方與賣方決定他們的偏好，包括關注之屬性、初始提案、權重、不同類型的效用函數 (utility function)或S型價值函數 (value function)。而過程中，雙方透過協商決策函數（negotiation decision function）之建議決定協商策略，以計算提案讓步的幅度並在有限的時間下促使共識達成。若未能達成共識，則買方繼續與第二順位之賣方進行同樣的協商流程，若所有的賣方皆無法與買方達成共識時協商結束。另外，我們首度將S型價值函數納入協商雙方的價值衡量，並運用統計檢定方法，證實價值函數與一般普遍使用的效用函數在成交提案與最佳提案之差距具有顯著差異，且價值函數結果優於效用函數。 The study aims to establish a web-based negotiation support system (NSS) with multiple attributes for mediating purchasing. The preferences of negotiators and the negotiation strategies and tactics are considered so that a well negotiation process is defined from a negotiation being initiated until an agreement being reached or broken down. The NSS can mediate firms, i.e., a buyer and vendors, to make a purchasing decision for durable goods effectively and efficiently. After a rank of vendors is provided from a group decision support system, the NSS provides a chance of mediation between the buyer and a vendor with the highest rank. Both can determine their preferences, including concerned attributes, weight of attributes, initial offers, and various types of utility functions or S-shaped value functions (in prospect theory) for representing adequate choices. In the process, both will also define their negotiation tactics by negotiation decision function for assigning the range of concessions so that an agreement can be reached with a limited time or offers. Otherwise, the system will continue to mediate the buyer and the vendor with the second highest rank through the same process. The negotiation might be broken down if all vendors'' do not match the buyer''s concerns. In addition, the preferences are described by the common utility functions and S-shaped value functions in the system. A small number of experiments are conducted through a statistical hypothesis testing, and the latter are shown a better performance.