風險態度之估計及其對猜題傾向的影響──以生物科選擇題為例

The Estimation of Risk Attitude and its Effect on Guessing Tendency: Taking a Multiple- Choice Biology Test as an Example

不少測驗會採用倒扣計分來抑制考生在選擇題的猜題行為。晚近研究結合決策理論與試題反應理論，發現在此計分方式下，考生風險態度會對其猜題傾向（答題行為）產生偏誤。但這些研究多不是實地研究，且經常忽略決策框架與學科能力對風險態度解釋力的影響。基於晚近研究的啟示與限制，本研究目的為：一、結合展望理論與試題反應理論，並用真實測驗資料，估計考生的風險態度與損失態度；二、應用本研究估計所得的風險態度及相關變項，探究不同決策框架下，學科能力如何調節風險態度對猜題傾向的解釋力。本研究以5,000位考生在大學指定考試生物科單選題的各題答題反應，做為估計其風險態度與損失態度的材料，發現本研究估計所得的風險態度與損失態度值，雖小於Tversky與Kahneman（1992）所估計者，但仍在其他相關研究的估計範圍內。此外，本研究也以上述5,000位考生中的3,765人為樣本，透過階層多元迴歸分析與Johnson-Neyman技術，進一步發現：在得框架題，學科能力愈高者，其風險態度對猜題傾向的解釋力愈大；在失框架題，學科能力愈低者，其風險態度的解釋力愈大。本研究的發現，不僅對考試實務工作深具啟示，也能應用於測驗與評量研究。

The ‘rights minus wrongs’ scoring method is often utilized to correct for guessing in multiple-choice items. Recently, by integrating decision theory with Item Response Theory, several studies have suggested that this scoring method may introduce guessing tendency (item response) biases through examinees’ risk attitude. However, none of these studies are field research and have investigated the effect of decision framing and academic ability on the prediction of risk attitude. Based on Prospect Theory and Item Response Theory, the first purpose of this study was to estimate the examinees’ risk attitude and loss attitude from high-stakes test data. By the estimated risk attitude and other variables, the second purpose of this study focused on how academic ability moderates risk attitude prediction about guessing tendency. The material for estimating risk attitude and loss attitude was 5,000 examinees’ multiple-choice item response data in Department Required Biology Test for college entrance. The estimates for risk attitude and loss attitude in the present study were lower than those Tversky and Kahneman (1992) estimated, but still in the interval that most related studies showed. For the second purpose of this study, 3,765 samples were drawn from the 5,000 examinees mentioned above, and their academic ability and guessing tendency were further estimated. To reveal the moderator effect of academic ability, hierarchical multiple regression and Johnson-Neyman technique were conducted in different decision frames. The results revealed that, in the gain frame, higher ability examinee’s risk attitude had more impact on guessing tendency. On the contrary, in the loss frame, lower ability examinee’s risk attitude had more impact. The findings in the study not only had far-reaching implications for examination practice but also could be applied to testing and assessment research.