協商(Negotiation)是常見的商業行為，但過程有時曠日費時，並需耗費許多人力物力。以軟體代理人為基礎的自動化協商，可有效降低總體協商成本，並加快成交速度。然而，自動化協商的普及有賴相關基礎建設的成熟，也與代理人所使用的協商策略是否符合協商者之實際需求息息相關。為此，本研究針對常用的自動化協商策略進行分析，以大規模的模擬協商來統計各種策略的效能。我們針對雙邊雙議題協商，使用加總計分模型(Additive Scoring System)[12]與Faratin的策略模型[3]進行分析。模擬實驗結果顯示: 一開始堅持、最後大幅讓步，而且完全不考慮對手反應的時間相依戰略戰(Boulware)可獲得最高的平均效用，當使用多種戰略來組合成策略時，含有Boulware之策略仍佔有優勢。行為相依戰略雖然在平均效用上不如Boulware，但在效率與公平指標上，卻遠超過Boulware戰略，有利於與對手建立長遠關係。此外，我們也發現時間相依戰略所產生的協商結果相當不穩定，不適合穩健的協商者。除了上述分析之外，我們也發展了一套協商支援系統，支援項目包括喜好擷取、策略支援與結果落點分析，以利協商者實際驗證各種策略的有效性，並藉此訓練協商技巧。

Negotiation is a common phenomenon in commercial business, but it takes lots of resource and effort. Agent-based automated negotiations can reduce resource and total cost. However, automated negotiations are not widely adopted because it depends on popularization of IT infrastructure and strategies that agents use during negotiations still can’t satisfy people. So, in this thesis, we study on the efficiency of automated negotiation strategies with simulated negotiations scenery. We analyze two-party, two-issue negotiations with Additive Scoring System[12] and Fartin's strategies model[3]. The result of our simulation shows that single Boulware tactic will reach the highest utility value and the compound tactics with Boulware still have some advantages of utility. Although behavior-dependency tactics are not as good as Boulware tactic on average utility, it far exceeds Boulware tactic on Equality and Efficiency index that can build a good long-term relationship with opponents. In addition, we observe that time-dependency tactics will easily reach unpredictable results, so it is not suit for negotiators want to avoid risk. Besides, we also develop a Negotiation Support System to help negotiators capture their preference, build strategies, settlement analysis and training. Finally, we will try to evaluate effectiveness of strategies with this system.

目錄
第1章	前言	1
第2章	協商相關理論簡介	5
2.1	協商分類(Taxonomy of Negotiations)	5
2.2	協商者的偏好(Preferences of Negotiators)	5
2.3	協商戰略與策略	7
2.4	可行解區域與效率前緣	11
第3章	模擬協商前置設定	13
3.1	協商案例(A Negotiation Case)	13
3.2	模擬設定	14
3.3	實驗結果評量指標	18
3.4	指標綜合評量（Overall Indexes, OI）	20
第4章	實驗結果與分析	21
4.1	平均效用值	21
4.2	與均等效率點的距離	27
4.3	效用值標準差	30
4.4	平均成交回合數	32
4.5	平均成交率	33
4.6	指標綜合評量	34
4.7	其他建議	36
第5章	協商支援系統	38
5.1	系統架構	38
5.2	Server端系統架構	39
5.3	Client端系統架構	41
第6章	結論與未來研究	44
6.1	結論	44
6.2	未來研究	45
參考文獻	47
附錄一  U模式各戰略組合之平均效用值	49
附錄二  I模式各戰略組合之平均效用值	53

表目錄
表2-1   單一議題之效用函數[11]	7
表3-1   協商案例之範例	13
表3-2   出價範圍100%重疊之喜好設定	14
表3-3   代理人a之提案表，以效用值排序	15
表4-1   採用U模式的平均效用值前五名	22
表4-2   只採用Boulware戰略時，表現較差的對戰組合	22
表4-3   採用U模式的平均效用值末五名	23
表4-4   I模式與U模式的平均效用值前五名比較	23
表4-5   I模式與U模式平均讓步值	25
表4-6   協商回合數與平均效用值的變化	26
表4-7   採用U模式的EE-dist前五名	28
表4-8   平均效用值前五名EE-dist比較	29
表4-9   各時間相依之戰略組合EE-dist值	30
表4-10  採用U模式的SDU排名	31
表4-11  採用U模式的ASRs排名	33
表4-12  平均成交率未達100%的戰略組合	34
表4-13  指標綜合評量權重設定	35
表4-14  依表4 13權重設定的OI排名	35
表6-1   自動化協商試驗表	45

圖目錄
圖2-1  時間相依的三種策略	9
圖2-2  雙方的共同可行解區域	11
圖3-1  既公平又有效率的成交點	19
圖4-1  使用U模式時，己方讓步造成的三種結果	24
圖4-2  Boulware與Boulware戰略組合的讓步過程	26
圖4-3  Boulware戰略讓步幅度隨著回合數的變化	27
圖4-4  U模式之協商過程	30
圖4-5  I模式之協商過程	30
圖5-1  系統架構圖	38
圖5-2  協商支援系統 Server 端架構圖	39
圖5-3  協商資訊圖	40
圖5-4  協商支援系統Client端架構圖	41
圖5-5  設定每一議題出價值、策略和戰略權重之畫面	42
圖5-6  協商畫面圖	43
圖5-7  Client端協商資訊圖	43

[1]Barbuceanu, M. & Lo, W.-K. (2000). “A Multi-Attribute Utility Theoretic Negotiation Architecture for Electronic Commerce,” Proceedings of the Fourth International Conference on Autonomous Agents, 239-246.
[2]Blecherman, B. (1999). “Adopting Automated Negotiation,” Technology in Society 21, 167-174.
[3]Faratin, P., Sierra, C. & Jennings, N. R. (1998). “Negotiation Decision Functions for Autonomous Agents,” Robotics and Autonomous Systems 24, 159-182.
[4]Foroughi, A. (1997). “Minimizing negotiation process losses with computerized negotiation support systems,” The Journal of Applied Business Research, Vol. 14, No. 4, 15-26.
[5]Goh, K.-Y., H.-H. Teo, H. Wu, and K.-K. Wei. (2000). “Computer-Supported Negotiations: An Experimental Study of Bargaining in Electronic Commerce,” Proceedings of the 21st International Conf. on Information Systems, 104-116.
[6]Kersten, G. E. & Noronha, S. J. (1999). “WWW-Based Negotiation Support: Design ,Implementation and Use,” Decision Support Systems 25, 135-154.
[7]Lee, C.-F., “Evaluation of Automated Negotiation Tactics in e-Business,” 資訊管理暨實務學術研討會, 2006, Taipei, Taiwan.
[8]Lim, J. (2003). “A Conceptual Framework on the Adoption of Negotiation Support Systems,” Information and Software Technology 45, 469-477.
[9]Lim, L.-H. (1999). “Multi-stage Negotiation Support: A Conceptual Framework,” Information and Software Technology, Vol. 41, 249-255.
[10]Lopes, F., Mamede, N., Novais, A. & Coelho,H. (2001). “Negotiation tactics for autonomous agents,” Proceedings of the 12th International Workshop on Database and Expert Systems Applications, 708-714.
[11]Mumpower, J. L. (1991). “The Judgment Policies of Negotiators and the Structure of Negotiation Problems,” Management Science 37 (10), 1304-1324.
[12]Raiffa, H.(1982). The Art and Science of Negotiation. Cambridge. MA: Harvard Univ. Press.
[13]Rangaswamy, A. & Shell, G. R. (1997). “Using Computers to Realize Joint Gains in Negotiation: Toward an Electronic Bargaining Table,” Management Science 43 (8), 1147-1163.
[14]Wang K. J. and C.H. Chou (2003). “Evaluating NDF-based Negotiation Mechanism within an Agent-based Environment,” Robotics and Autonomous Systems 43, 1-27.
[15]Wurman, P. R., M. P. Wellman, and W. E. Walsh. (1998). “The Michigan Internet AuctionBot: A Configurable Auction Server for Human and Software Agents,” Proceedings of Second International Conference on Autonomous Agents, 301-308.
[16]李青芬(2006)，電子化協商之期望效用、戰略及結果預測之研究，國立交通大學，經營管理研究所博士論文。