決策不確定效應分流機制之模型建立

Modeling Streamlining of Ambiguity Effect in Decisional Process

本研究旨在檢證不確定效應的分流機制。當決策者對於事件發生可能性估計感到不確定時，該不確定感分別影響到決策者對事件機率估計的調整（P機制，分流一），以及對心理代價評估的調整（U機制，分流二）。這種不確定感經由該兩種途徑以影響決策與選擇的偏好，稱為不確定效應的分流機制。本研究提出不確定效應存在P與U機制的分流，而且指出該雙機制皆由不確定程度與態度決定效應的大小與方向，並進而發展表現該雙機制的函數模型。本研究設計兩組實驗，分別就不同決策作業（訴訟談判／股市預測），操弄不確定感來源與程度（證據份量／領域知識），以及不同不確定態度（決策作業特性／決策者認知類型）。實驗結果就分流機制特性所推導的下列假設加以驗證：（1）存在不確定感程度效應；（2）存在不確定態度的效應；以及（3）存在不確定分流機制。結果顯示上述假說皆得到兩組實驗結果的一致支持，驗證了不確定效應分流假說。並且（4）所發展Liou模型較其他雙參數模型在表現P機制與U機制上具備最佳函數配適性，並且該模型之參數與分別在表現該雙機制之「不確定感程度」與「不確定態度」兩種心理效應，具備最佳參數靈敏性。

The present study tries to validate the dual mechanisms of ambiguity effect in decisional process. Dual mechanisms operate in a streamlining way when a decision maker is uncertain of the occurrence probability of certain events, The ambiguity thus incurred modulates the estimation of event probability (P-mechanism or Streamline 1) and that of event consequence (U-mechanism or Streamline 2). We propose the existence of the streamlining of dual mechanisms, and argue that these two mechanisms are both influenced by perceived degree of ambiguity and attitude toward ambiguity. Furthermore, we developed a mathematical model to express these mechanisms. Based on the proposed streamlining mechanisms, two experiments were conducted to test three hypotheses derived from the revised model: (1) existence of the 'degree of perceived ambiguity' effect; (2) existence of the 'ambiguity attitude' effect; and (3) existence of the streamlining of dual mechanisms. The experimental results are in general agreement with the three hypotheses. Our demonstrations thus clearly complement the single-mechanism (i. e., P or U-mechanism, but not both in a streamlining way) explanations in the literature. The present study follows a substantial revision of Lattimore et al's (1992) model to separate and to accommodate the effects of 'degree of perceived ambiguity' and 'ambiguity attitude'. The revised Liou model (Liou, 1998) was then used to simulate the behavior of P- and U-mechanisms under various combinations and values. By comparing Liou model with other ambiguity models in da- ta-fitness and parameter sensitivity, which proposed by Einhorn & Hogarth (1985), Kahn & Sarin (1988), and Lattimore et al (1992). The test results showed that Liou's model is the best functional form in modeling P and U mechaism.