動態政策環境中的決策輔助：標竿學習的實驗研究

Abstract

許多公共議題的政策擬定、執行、評估、與修正本質上是一個典型的動態決策環境，有限理性的公共管理者的職能即在此動態政策環境中制定決策。據此，本研究提出一理論模型以作為設計動態政策環境下決策輔助的基礎，在一個電腦模擬的實驗中，本研究將於兩個不同的動態複雜度下驗證三個作為決策輔助的成果回饋。 實驗結果顯示:具備標竿成果回饋資訊的受測者在績效及學習方面均優於只有一般成果回饋的受測者；但是當同時提供標竿成果及標竿決策時，反而對受測者在續效及學習兩方面皆有不良影響甚至劣於只有一般成果回饋的受測者，透過分析所提出的理論模型中的因徑後發現，此不良影響主要經由受測者選擇資訊陳列格式的行為而發生作用。此研究發現除了提供關於標竿學習的進一步實證外，對建立政策實驗室也具有實質的貢獻。

Public policy-making can be well approached as human decision-making behavior striving to manage complex dynamic tasks. Particularly, various governmental programs require public officials make decisions on a regular basis with the next decisions depending on the outcome of previous decisions. This study proposes alternative designs of outcome feedback and hypothesizes that they help decision-makers manage dynamic decision tasks more effectively. A process model describing how the relevant components of dynamic decision-making behavior - task difficulty, information display, task knowledge and task performance - interact with one another is also theoretically proposed and empirically examined. The empirical evidence supports the hypothesis that providing benchmark outcome improves task performance. Path analyses show that the improvement comes from improving the anchoring and adjustment heuristics the subjects used. The findings contribute to how decision support may be designed to aid decision-making in complex dynamic systems in general and public policy-making laboratory in particular public policy-making laboratory in particular.