不確定下的決策行為主要判斷事件發生機率與事件後果。近來實徵研究一致指出，決策加權函數(Decision Weighting Function)違反機率運作原理，而呈現反S型的非線性心理物理轉換。Tversky與Wakker(1995)提出「有界的次加成」特徵與測量描述這種以機率1或0為參考點的敏感遞減現象。Gonzaiez與Wu(1999)繼之分別以「區辨性」與「吸引性」說明該函數的曲線「曲度」以及「升降」的心理物理函數特徵。本研究一以表現該兩項特性為判準。比較不同決策加權函數(分別為Einhorn & Hogarth, 1985; Kahn & Sarin, 1998; Lattimore et al., 1992)的描述力，並提出實徵特性較佳的雙參數Liou模型。研究二即對該兩項特徵提出心理解釋，假設「不確定感程度」(感覺因素)影響決策加權函數的區辨性，而「不確定態度」(非感覺因素)則影響函數的吸引性。在分別操弄「不確定感程度」(証據份量及熟悉程度)與「不確定態度」(得失觀點及後果嚴重性)，結果証實所提出的解釋。並經由研究二實徵資料的曲線符合度與敏感度分析，証實Liou模型為最具描述力的決策加權函數型式。函數(Utility Function，以U(X)表示)轉換而得的效益值(U(Xi))表示。而某個選項的吸引力(SEU值)，即是各種後果的效益值乘以主觀機率(P(Xi))權重後的加成而時，即SEU(X)= 。後繼理論以反映決策選擇行為真實性的描述取向，即注重效益與主觀機率的心理轉換而修正SEU理論(Kahneman & Tversky, 1979; Tversky & Kahnema, 1992)。 近來實徵研究一致指出，決策者對機率的估計反機率運作原理，而呈現反S型的非線性心理物理轉換，稱為機率加權函數(Probability Weighting Function)。亦即機率加權函數呈現低機率高權重、高機率低權重，先可後臺的反S型曲線特性。機率加權並非可由主權機率直接預測，而是反應決策者機率運作的心理物理轉換(Gonzales & Wu, 1999)，故亦稱之為決策加權函數(Decision Weight Function，例如Tversky & Kahneman, 1992)。若出A後果的機率為P(P，A)，則決策加權(Decision Weight，簡稱DW)與P的對應函數，DW=f(P)，f(‧)即為決策加權函數。一般DW值的導出乃採「確定等價法」(Certainly Equivalence，簡稱CE)。此法要求決策者對(P，A)等值出價為CE，則U(CE)=DW(A)．U(A)，可得DW(A)=U(CE)/U(A)。

Empirical studies have shown that the probability weighting function is in an inverse S shape, which is concave for the low probability range and convex for the high probability range. Gonzalez and Wu (1999) proposed discriminability (involving the curvature of a function) and attractiveness (involving the elevation of a function) as two distinctive features of the weight function. This research aims to examine factors that influence these two independently psychological properties. We proposed that the degree of perceived ambiguity influences probability discriminability by diminishing sensitivity, and ambiguity attitude influences attractiveness that modulates the elevation of a weighting function. Applying these two features as criteria, Study 1 tested three different functional forms of decision weight proposed by Einhorn and Hogarth (1985), Kahn and Sarin (1988), and Lattimore, Baker, and Witte (1992). Results showed that these three models parameter values. By reparameterizing the function of Lattimore et al., we developed the Liou model which claracterizes decision weight with two parameters such that one parameter Θrepresents the curvature (discriminability) and the other Pc represents the elevation (attractiveness). Using parameter values of the Liou model as indicators, Study 2 manipulated the degree of perceived ambiguity by evidence weight (relative amount of missing information) in Experiment 1, and feeding of competence (expertness and novice of the decision task) in Experiment 2. Ambiguity attitudes were manipulated by gain or loss perspective in Experiment 1, and by seriousness of consequence in Experiment 2. results of these two experiments convergently supported the hypothesis that the degree of perceived ambiguity modulates probability discriminability which shapes the curvature of a decision weight function, and that the ambiguous attitude of decision makers influences attactiveness of the choice task which modifies elevation of the function. Based on empirical data of Study 2, we then tested curve fitness and analyzed the sensitivity of models discussed in Study 1. Results indicated that the Liou model fits best and is most sensitive in representing the decision weight function. Implication of these findings for psychological significance of decision weight function is discussed.