當通過注意力檢測題需要的不只是注意力：加入誘餌題的結果

When It Takes More Than Attention to Pass an Attention Check: Effect of Including a Lure Question

網路填答者若答錯注意力檢測題，會被認定是不認真。但檢測題必須能正確區分有用心跟不用心者，才具效力。本研究提出，若檢測題不但指定特定的答案，並且包含一個誘餌題，則該檢測題不但通過率低，效力也降低。其原因是，填寫者必須花額外的心力才能放棄自己在誘餌題的回答，但很多認真填寫者沒有多花這額外心力，而以自己的回答作答，進而被認定答錯。此外，因為要花額外心力，有通過這類檢測題的人，會比沒通過的人花費明顯較多的時間；相對地，沒有誘餌題的檢測題，不論答對答錯都不需要花那額外心力，所以兩者在填答時間上的差異相對小。
以上論述以四個假設檢驗。結果發現，被認定是錯的答案，明顯是在回答誘餌題。有誘餌題時檢測題的通過率及效力都比較低，沒有誘餌題時比較高。但有關額外時間的假設，只在一個實驗得到支持。結論：檢測題若指定填寫內容但不包含誘餌題，效力最佳。

Attention checks (ACs) are inserted in survey questionnaires to differentiate between attentiveness and inattentiveness. However, some ACs result in low passing rates. This study suggests that ACs with an instructed response and a lure question (LQ) result in lower passing rates and effectiveness than ACs without an LQ. It argues that respondents must make an extra cognitive effort to disregard their answers to the LQ. Consequently, even attentive respondents can be flagged as inattentive if they do not expend the extra effort by providing their answers to the LQ. Then, flagged responses to such ACs tend to be answers to the LQ (Hypothesis 1), and their passing rates are lower than those of ACs without an LQ (Hypothesis 2). More¬over, attentive respondents are likely to produce high-quality data, but they are more likely to be flagged by ACs with an LQ than ACs without. As a result, the difference in data quality between passers and failers of the former will be smaller than that of the latter. That is, the effectiveness of ACs with an LQ is lower. (Hypothesis 3). Additionally, because extra cognitive effort requires additional time, the response time difference between those who pass and fail ACs with an LQ is larger than that of ACs without an LQ (Hypothesis 4).
Hypotheses 1, 2 and 3 were supported. Hypothesis 4 was supported by only one experiment. This study discusses the implications, limitations, and directions for future research.