The Ideal Point Model in Action: How the Use of Computer Adaptive Personality Scales Benefits Organizations

Abstract

In their final thoughts, Drasgow, Chernyshenko, and Stark (2010) state that although the ideal point model holds much promise for improved assessment instruments, there are relatively few applications of the approach. We and more than 50 scientists from Personnel Decisions Research Institutes, PreVisor, and Navy Personnel Research, Studies, and Technology have applied the ideal point assumptions of the item response process to the development of a computer adaptive testing (CAT) approach to the measurement of personality traits. On the basis of the algorithm underlying the computer adaptive rating scales method (CARS;Borman et al., 2001), this approach leverages advances in measurement and technology to efficiently assess employees' traits and characteristics. These efforts have led to the development of new personality instruments for use in the military and private sectors. In this commentary, we discuss the practical benefits of CATbased personality measurement and identify areas for further research inquiry.

Computerized Adaptive Testing with the Zinnes and Griggs Pairwise Preference Ideal Point Model

“…In addition, pairwise preference tests are attractive in situations where the pool available for test construction is relatively small because statements can be used in multiple pairings to create a much larger number of potential items. The desire to use pairwise preference items is also reflected in the relatively recent appearance of computerized adaptive tests for applications, such as performance appraisal (CARS II; Borman et al, 2001;Schneider et al, 2003;Stark & Drasgow, 2002) and personality assessment (Houston et al, 2005;Kantrowitz & Tuzinski, 2010). However, relatively little research has been published about the psychometric properties of pairwise preference models, much less how they can be used as the basis for CAT.…”

Fitting item response unfolding models to Likert-scale data using mirt in R

“…Unfolding models are suitable when the underlying measurement process contains a proximity property with respect to the item-level stimuli. Although they have attracted huge attention recently [ 9 , 12 , 15 , 20 , 64 ], the development of parameter estimation software for various unfolding models has largely been left behind. To enhance the utilities of unfolding models in practice, the mirt package was adopted in this article and evaluated using Monte Carlo simulation studies.…”

Improving the Measurement of Psychological Variables: Ideal Point Models Rock!

“…Kantrowitz and Tuzinski (2010) describe two personality CATs, the Navy Computer Adaptive Personality Scales and the GPI‐Adaptive, that use the same underlying ideal point model (the Zinnes and Griggs' [1974] model) as CARS. They report uncorrected validities of composites that exceed .30, which has long been viewed as an upper limit.…”

“…It seems reasonable to assume that the levels of performance associated with the ratee and the two statements are perceived with error, which adds to the complexity of the psychometric model needed for CARS. Kantrowitz and Tuzinski (2010) describe two personality CATs, the Navy Computer Adaptive Personality Scales and the GPI-Adaptive, that use the same underlying ideal point model (the Zinnes and Griggs' [1974] model) as CARS. They report uncorrected validities of composites that exceed .30, which has long been viewed as an upper limit.…”