發現交互作用困難嗎？－－「探究目標」對於大學生發現變項間交互關係的影響

Improving Your Chance to Discover an Interaction Effect by Changing the Inquiry Goal

能判斷事物之間的交互作用效果對科學探究或日常因果推論至為重要，但科學推理的研究顯示不論兒童或成人很少能主動發現交互作用效果，過去研究者多認為這是因為人們缺乏相關概念或推理能力所致。本研究以三個實驗檢測大學生參與者在自主實驗作業（self-directed experiment task）中找到交互作用的比率以及其探究歷程是否因「探究目標」不同（找出會影響某結果的原因或找出能預測結果的因果規則）而改變。結果顯示在三個不同因果結構複雜度不同的探究作業中（實驗一至三），因果規則組皆測試更多交互作用假設，更少單一變項假設，也有更高的交互作用發現率，顯示探究目標會影響推理者交互作用假設的可得性與檢驗傾向，進而影響交互作用發現的機會。此一影響即便在容易得到所有實驗組合資料的情況下（實驗三）仍存在。本研究首次發現交互作用的探究會受到與經驗內容無關的情境脈絡因素影響，不但意涵過去研究有可能低估成人發現交互作用的能力，也揭示提升探究表現的途徑，對科學教育有重要意涵。

Previous research on scientific reasoning has shown that adults and children are poor at discovering interaction effects in inquiry tasks that involve multi-cause structures. Lack of a relevant causal framework and relevant reasoning skills have often been proposed to explain the failure rate. In this study we aimed to demonstrate that participants' performance finding interaction effects could be greatly improved by changing the inquiry goal even when the relevant abilities and task were kept the same. Three experiments with a self-directed experiment (SDE) paradigm were conducted to test the hypothesis at different levels of inquiry task complexity. With the SDE paradigm, participants could change the values of candidate variables in each test trial and observe the effect (i.e., the speed of toy racing cars) in order to figure out the underlying causal structure. Undergraduate participants were randomly assigned to two groups that only differed by the inquiry goal. In the causal-status group, particip ants were instructed to determine which variables could influence the effect of concern and how, as previous studies usually did; in the causal-rule group, participants were instead told to find a causal rule to predict the effect. In addition, participants' abilities at inferring an interaction effect from a set of double-controlled experiments (changing the value of a variable at each level of the other variable while keeping other variables constant) and designing experiments to test a target interaction hypothesis were also measured. As predicted, the success rate at finding a target interaction between two candidate causes for casual-rule group was twice as much as for casual-status group in all three experiments. The causal-status group performed poorly as previous studies reported, even though their abilities of inferring and testing interactive hypothesis were as good as the other group's. Moreover, the causal-rule group tested more interactive hypotheses, less single-variable hypotheses, and thus had more chances to gather evidence about interaction with the same amount of testing trials. These results, for the first time show that adults' performance on finding interactive effects can be influenced by a top-down contextual factor, i.e., the inquiry goal, via changing the availability of interactive hypotheses in the reasoner's mind. Implications for science education are discussed.