Age-related differential item functioning in tests of face and car recognition ability
Age-related differential item functioning in tests of face and car recognition ability
复制标题
年龄差异项在人脸、汽车识别能力测试中的作用
DOI：
10.1167/18.1.2
复制
发表时间：
2018
影响因子：
1.8
通讯作者：
Gauthier, Isabel
中科院分区：
文献类型：
--
作者：
Sunday, Mackenzie A.;Lee, Woo-Yeol;Gauthier, Isabel
测试中项目功能差异(DIF)的存在表明存在偏见，这可能对特定群体的成员不利。之前关于视觉学习能力测试的研究发现，在汽车测试中，DIF与年龄组显著相关(Lee，Cho，McGugin，Van Gulick，&Gauthier，2015)，但在面孔测试中不存在(Cho等人，2015)。年龄DIF的存在对测试的有效性构成了威胁，即使对于那些对衰老不感兴趣的研究也是如此。在这里，我们评估了这种针对汽车而不是面孔的年龄DIF模式是否也适用于针对相同能力的新测试，该测试采用了一项新的匹配任务，每次试验使用两个研究项目。我们在人脸和汽车的匹配测试中找到了DIF的证据，尽管效果大小令人欣慰。尽管年龄DIF在测试水平上足够小，对大多数用途来说是可以接受的，但我们也询问了我们匹配任务的特定格式是否会导致一些与年龄相关的DIF，而不考虑领域。我们将面孔匹配任务分解为其组成部分，并使用执行这些更简单任务的受试者的新数据，发现了年龄DIF是由每次试验研究中呈现的两张面孔的相似性驱动的证据。总体而言，我们的结果表明，使用匹配格式，特别是对于汽车，可以减少与年龄相关的DIF，并且每个试验只有一个研究项目的更简单的匹配任务可以进一步减少年龄DIF。
The presence of differential item functioning (DIF) in a test suggests bias that could disadvantage members of a certain group. Previous work with tests of visual learning abilities found significant DIF related to age groups in a car test (Lee, Cho, McGugin, Van Gulick, & Gauthier, 2015), but not in a face test (Cho et al., 2015). The presence of age DIF is a threat to the validity of the test even for studies where aging is not of interest. Here, we assessed whether this pattern of age DIF for cars and not faces would also apply to new tests targeting the same abilities with a new matching task that uses two studied items per trial. We found evidence for DIF in matching tests for faces and for cars, though with encouragingly small effect sizes. Even though the age DIF was small enough at the test level to be acceptable for most uses, we also asked whether the specific format of our matching tasks may induce some age-related DIF regardless of domain. We decomposed the face matching task into its components, and using new data from subjects performing these simpler tasks, found evidence that the age DIF was driven by the similarity of the two faces presented at study on each trial. Overall, our results suggest that using a matching format, especially for cars, reduces age-related DIF, and that a simpler matching task with only one study item per trial could reduce age DIF further.