Computer-Adaptive Assessment of Temperament to Support Personnel Selection and Classification Decisions

We intend to use our recently developed IRT model for scoring multidimensional pairwise preference items as the basis for a computerized adaptive test (CAT) that is designed to be fake resistant. The items will consist primarily of pairs of personality statements, on different dimensions, which are similar in social desiriability. A respondent's task will be to choose the statement in each pair that is "more like me." Our recent studies have shown this IRT approach to constructing and scoring fake-resistant items can successfully recover normative scores when at least two personality dimensions are assessed. Moreover, the use of adaptive item selection should greatly facilitate the process of test construction, increase measurement accuracy, and decrease testing time. In Phase I, we will 1) conduct a series of simulation studies involving our IRT approach to constructing and scoring multidimensional pairwise preference items; and 2) identify important and nonoverlaping temperament dimensions, relevant to performance in broad military and civilian job groups.