證據取樣說對後見之明偏誤的解釋力

THE EXPLANATORY POWER OF THE EVIDENCE-SAMPLING HYPOTHESIS FOR HINDSIGHT BIAS

後見之明偏誤是指個體在得知事情的結果之後，會拒絕結果訊息的影響而高估自己能正確預測事情發生機率的現象。以往研究在各種不同的範疇中證實了後見之明偏誤現象的普遍性，數位學者也提出相關機制來解釋後見之明偏誤，但較少研究者從驗証後見之明偏誤的內在機制之角度切入。本研究的主要目的即是想藉由分析受試者的內在想法，來評估證據取樣假說對後見之明偏誤的解釋力。本研究採用後見之明偏誤的受試者內實驗設計研究典範，實驗為一個單因子（結果：有／無）受試者間設計。研究結果顯示，受試者接受了結果訊息後確實產生了後見之明偏誤，但不論受試者有、無接受結果訊息，其提取出與結果一致想法的比例並未達顯著差異，因此本研究結果並不支持證據取樣假說對後見之明偏誤的解釋。後續內文將對可能的問題和可能的內在機制做進一步的釐清和討論。

Hindsight bias refers to the phenomenon that, upon learning the outcome of an event or the answer to a question, individuals tend to inflate what could have been predicted in foresight. Numerous studies have demonstrated the phenomenon of hindsight bias and its ubiquity. Several underlying mechanisms (e.g., memory-change hypothesis, evidence-sampling hypothesis, anchoring and adjustment, and motivation) have been proposed for such a phenomenon. However, only a few of hindsight bias studies investigated the explanatory power of these mechanisms. After comparing these mechanisms, the authors find that the evidence-sampling hypothesis is more concrete and provides an index of outcome-consistent thought ratio for further testing of the hypothesis. The main purpose of the present study, thus, is to investigate the explanatory power of the evidence-sampling hypothesis by analyzing the subject's spontaneous thought listing. Ninety-two undergraduate students participated in the two-session experiment. The experiment, adopting a within/subjects hindsight bias research paradigm, was a one factor (outcome knowledge: yes/no) between/subjects design study. In Session l, the subject was asked to choose the correct answer for each question from two alternatives, then to state the likelihood that the chosen answer was indeed correct. In session 2, the subject was informed of, for each question, the correct answer and was asked to recall the chosen answer and the probability reported in Session I. The subject was further asked to write down any thoughts that influenced his/her current answer choice. After some data transformations, the mean magnitude of hindsight bias and the mean outcome-consistent thought ratios were computed separately for the outcome group and no outcome group. The results showed that subjects with outcome knowledge indeed exhibited hindsight bias. Nevertheless, there were no differences in the outcome-consistent thought ratios between subjects with and without outcomes. The results of the present study do not support the evidence-sampling hypothesis for hindsight bias. Some possible problems and underlying mechanisms are further discussed.